Daily Tech Digest - May 27, 2026

---

Quote for the day:

“If you can get today’s work done today, but you do it in such a way that you can’t possibly get tomorrow’s work done tomorrow, then you lose.” -- Martin Fowler

---

🎧 Listen to this digest on YouTube Music

▶ Play Audio Digest

Duration: 23 mins • Perfect for listening on the go.

CERT-In’s new AI cybersecurity blueprint urges 12-hour remediation for known exploited vulnerabilities

India’s cybersecurity regulator, CERT-In, has released a 38-page guideline addressing the growing risks of artificial intelligence in cyberattacks. The document details how adversaries are using automated tools to speed up data collection, phishing, and malware creation, which severely shortens the time organizations have to defend themselves. To combat this, the regulator recommends that enterprises patch, isolate, or mitigate any known exploited vulnerabilities on critical internet-facing systems within twelve hours, while other major external flaws should be resolved within a single day. Because traditional methods like periodic audits and static defenses are too slow for rapid threats, the report encourages businesses to shift toward continuous system monitoring and automated response management. Beyond external threats, the text addresses internal risks within corporate environments, warning against employee use of public AI platforms that can leak sensitive data. It stresses the necessity of structured governance and human oversight over autonomous software decisions. Furthermore, the regulator explicitly reminds organizations of their mandatory statutory obligation to report all cybersecurity incidents within six hours. Ultimately, the document highlights that managing modern network risk is no longer just about establishing static defenses, but about responding quickly enough to isolate threats before automated attackers can completely outpace human security teams.

Why data governance is a core IT responsibility in the AI era

The article outlines why data governance has shifted from a routine compliance exercise to a primary responsibility for information technology teams in the era of artificial intelligence. Traditional data management handled structured tables, but modern systems consume vast amounts of unstructured information, such as emails, documents, and chat records. When internal company files are fed into modern automation tools and language models, any hidden errors or biases become heavily amplified. Because these automated software programs query data continuously and lack human skepticism, they process flawed inputs without question, turning upstream data failures into widespread operational errors. To address this, technology leaders must avoid common pitfalls like relying strictly on software purchases to patch broken processes or treating data strategy as a one-time project. Instead, a practical and sustainable approach requires close, cross-department collaboration with legal, risk, and business units to build a unified system for tracking data origins and real-world meaning. Rather than attempting to catalog every single file all at once, organizations should prioritize documenting and continuously monitoring their most high-impact information assets. Ultimately, treating corporate data as a carefully managed strategic resource ensures that underlying inputs remain strictly accurate and reliable, providing a dependable foundation for safe, effective, and predictable digital tools.

Responding to Breaches With AI? Beware Cross-Contamination

The article outlines important warnings for cybersecurity investigators who utilize artificial intelligence tools to draft incident response reports. Based on controlled experiments by Cisco's threat intelligence group, Talos, researchers found that large language models are highly susceptible to data cross-contamination. When multiple security incidents are processed during a single conversation session, information from a previous report can easily bleed into a subsequent one. Surprisingly, this data mixing occurs even if investigators completely delete the notes from the earlier incident before starting the next file. This core issue stems from the finite memory constraints of an AI's fixed context window, which often leads to unpredictable data blending as the conversation continues. Producing inaccurate reports introduces significant professional, regulatory, and legal liabilities, especially for multi-tenant incident response firms handling private customer data. Furthermore, the Talos tests revealed that models often deliver entirely inconsistent recommendations when fed identical data. To address these technical limitations, researchers recommend opening entirely new sessions for separate investigations and using structured prompting strategies. Breaking tasks into narrow instructions, enforcing rigid formatting templates, and specifying exact source documents cut down overall drafting time by half while minimizing errors. Ultimately, human oversight remains vital to catch hallucinations and guarantee report accuracy.

5 Security Principles Every Entrepreneur Should Apply to Leadership

In an essay published on APMdigest, Prakash Mana explains how the core principles behind cybersecurity offer a highly practical guide for business leadership. Rather than focusing purely on technical tools like network firewalls or data encryption, the author suggests that entrepreneurs can use these structural concepts to better manage risk, organizational trust, and long-term stability. The first approach involves adopting a continuous verification mindset toward trust, meaning that effective leaders stay curious and validate their strategic assumptions rather than relying blindly on company hierarchy or past achievements. Second, applying the standard security rule of giving the lowest level of privilege needed helps founders delegate responsibilities with clear, distinct boundaries, matching decision rights to specific expertise to prevent both micromanagement and employee burnout. Third, instead of allowing single points of failure to threaten the company, resilient businesses build multiple layers of protection by using cross-trained teams and clear, written operational routines. Furthermore, prioritizing open visibility over rigid control allows executives to address problems early and cultivate an environment of safety, rather than leading through heavily filtered corporate reports. Ultimately, the piece argues that borrowing these foundational practices helps leaders make calm, balanced choices in unpredictable market conditions, creating durable companies designed to grow steadily over time.

Digital Bank Employees Used to be the Stuff of Science Fiction. Not Anymore

The article from The Financial Brand examines how conversational and generative artificial intelligence systems are transitioning from theoretical concepts into practical workforce realities across the banking sector. Rather than replacing traditional core platforms or forcing a massive overhaul of human talent, modern artificial intelligence is primarily functioning as sophisticated middleware. Financial institutions are integrating task-specific digital assistants directly on top of decades-old back-office systems to streamline repetitive operational tasks. Major institutions like Morgan Stanley, Citigroup, and BNY Mellon have deployed knowledge management layers and multimodal systems that safely analyze text, voice, and documentation without disrupting strict regulatory standards. Similarly, smaller entities such as Grasshopper Bank have enabled business customers to securely link their accounting data directly to intelligent tools for automated reporting and immediate insights. This transition emphasizes a broader shift toward operational support and administrative efficiency, specifically targeting complex procedures like fraud prevention, compliance reviews, and transaction reconciliations. By taking over high-volume administrative drudgery, digital employees allow human personnel to focus on client relationships and complex problem-solving. This shift marks a practical, evolutionary upgrade rather than a radical disruption of the financial ecosystem.

Closing the Gap Between Security Ambition and Operational Reality

The article outlines the persistent friction between an organization's high security goals and its daily operational constraints. Many well-intentioned security updates inadvertently backfire by introducing excessive complexity, turning vital protections into frustrating bottlenecks for development teams. This issue usually surfaces when newly introduced security tools clash with established engineering workflows and fragmented old systems, forcing staff to spend valuable time manually tracking down alerts across multiple separate dashboards. To fix this common disconnect, the author argues that sustainable security excellence depends entirely on a foundation of solid operational maturity. Successful organizations achieve this stable state by utilizing modern cloud architecture that reduces unnecessary systemic complexity, using automation to eliminate repetitive manual tasks, and fostering a supportive team culture grounded in blameless problem solving. Instead of forcing unrealistic or overly aggressive timelines onto software engineering teams, which can take up to four years to successfully complete in highly complex environments, leaders should prioritize strengthening their core workflows first. Using gradual and incremental strategies to phase out outdated platforms allows companies to maintain steady protective coverage over time. This patient, methodical approach ensures that security measures naturally support day to day software development rather than obstructing it.

The Two Concepts Every Architect Needs to Master

In this article, Paul Preiss of Iasa Global outlines how architectural teams can take a structured, realistic approach to assessing business projects by using two collaborative tools from the Business Technology Architecture Body of Knowledge framework. Instead of relying on traditional timeline roadmaps, Preiss advocates for a team process that combines the Business Case Canvas and the Strategic Roadmap Canvas as active, shared working surfaces. The process begins with building an individual business case for each new proposal using the NABC format, which requires evaluating its true business need, specific technical approach, qualitative and quantitative benefits, and complete lifecycle costs. Once these criteria are established, the roadmap canvas allows business, solution, and technical architects to collectively evaluate proposals across key dimensions like value, structural complexity, regulatory compliance, and alignment with foundational principles. To prevent senior or vocal team members from inadvertently skewing the results, the team uses an independent, simultaneous scoring protocol that highlights conflicting perspectives early on. Finally, technical architects map out strict structural dependencies to determine the logical order of project execution. By unifying these insights, the architecture community develops an honest picture of organizational demand, moving funding debates away from office politics and toward clear, balanced investment conversations with business stakeholders.

Embracing an Offensive Mindset in Proactive Risk Management

The Disaster Recovery Journal article discusses how moving from a reactive stance to a proactive, forward-looking strategy improves organizational security. Traditional risk management usually addresses problems only after they happen, which frequently leaves companies highly vulnerable to unpredictable or sophisticated threats. To address this exposure, the author highlights the clear value of adopting an offensive mindset, where security teams actively look for hidden weaknesses before they can be exploited. This systemic transition requires a structured framework that starts by securing executive support and building an internal workplace culture where all employees feel genuinely responsible for pointing out potential hazards. Next, organizations must collect reliable internal data and external threat intelligence to gain full visibility over their digital and physical operations. Operational teams then set clear protocols to carefully evaluate and prioritize these findings based on their potential business impact. Finally, teams conduct structured threat hunts and cooperative exercises to continually test their defenses. This strategy shifts safety measures from a simple cost center to a core driver of stability and performance. By identifying internal flaws early and establishing a continuous feedback loop, companies can better safeguard their staff, secure sensitive data, and maintain steady operations over time.

Connected vehicles, disconnected security: Why connectivity architecture now matters most

Modern vehicles have essentially become computers on wheels, with hundreds of millions of connected cars currently driving on our roads. By the end of this decade, a single typical vehicle is expected to generate 25 gigabytes of data every hour. This massive volume of information travels across a mix of public and private networks, often without clear oversight regarding how it is routed or where it might be vulnerable. Historically, security strategies focused on protecting specific software applications or devices, assuming the communication paths between them were secure. However, because modern vehicle data moves through dozens of separate and uncoordinated routes, those traditional assumptions are no longer safe. To solve this problem, companies are changing their approach by treating the network architecture itself as the main foundation for security. Instead of relying on the public internet or open interconnections, they are setting up controlled exchange points to get better visibility and apply rules consistently. Ultimately, vehicles are no longer standalone products; they are pieces of a much larger, distributed system. Keeping them safe requires looking at the paths data takes and understanding how a failure in one area can ripple through the entire network.

Beyond the Org Chart: Why Your SRE Team Needs a Membrane, Not a Silo

In this article, a site reliability engineering leader shares how their department successfully resolved a severe operational crisis after multiple company acquisitions caused routine, repetitive maintenance tasks to consume nearly eighty-four percent of their overall workload. Instead of building a rigid, isolated silo that cuts off communication or leaving their doors wide open to an overwhelming firehose of incoming requests, the team introduced the concept of an organizational membrane. This semi-permeable boundary uses carefully calibrated triage criteria on intake boards to filter incoming assignments. Such a strategy successfully protects engineers from distracting daily noise while ensuring that genuine, high-priority system requirements still pass through. By treating the entry boundary as a serious engineering problem to be solved systematically rather than merely dismissing it as soft administrative work, the team drove their repetitive task ratio down significantly to under forty-five percent. Furthermore, they managed to shorten their task turnaround times significantly, dropping their longest completion cycles from two hundred ninety-four days down to just fifty-seven days. Ultimately, the author shows that implementing a thoughtful intake process allows internal operations teams to stay collaborative and helpful to the broader company without sacrificing their core focus on long-term system stability and software reliability.

Daily Tech Digest - September 09, 2025

---

Quote for the day:

“The greatest leader is not necessarily the one who does the greatest things. He is the one that gets the people to do the greatest things.” -- Ronald Reagan

---

Neuromorphic computing and the future of edge AI

While QC captures the mainstream headlines, neuromorphic computing has positioned itself as a force in the next era of AI. While conventional AI relies heavily on GPU/TPU-based architectures, neuromorphic systems mimic the parallel and event-driven nature of the human brain. ... Neuromorphic hardware has shown promise in edge environments where power efficiency, latency and adaptability matter most. From wearable medical devices to battlefield robotics, systems that can “think locally” without requiring constant cloud connectivity offer clear advantages. ... As neuromorphic computing matures, ethical and sustainability considerations will shape adoption as much as raw performance. Spiking neural networks’ efficiency reduces carbon footprints by cutting energy demands compared to GPUs, aligning with global decarbonization targets. At the same time, ensuring that neuromorphic models are transparent, bias‑aware and auditable is critical for applications in healthcare, defense and finance. Calls for AI governance frameworks now explicitly include neuromorphic AI, reflecting its potential role in high‑stakes decision‑making. Embedding sustainability and ethics into the neuromorphic roadmap will ensure that efficiency gains do not come at the cost of fairness or accountability.

10 security leadership career-killers — and how to avoid them

“Security has evolved from being the end goal to being a business-enabling function,” says James Carder, CISO at software maker Benevity. “That means security strategies, communications, planning, and execution need to be aligned with business outcomes. If security efforts aren’t returning meaningful ROI, CISOs are likely doing something wrong. Security should not operate as a cost center, and if we act or report like one, we’re failing in our roles.” ... CISOs generally know that the security function can’t be the “department of no.” But some don’t quite get to a “yes,” either, which means they’re still failing their organizations in a way that could stymie their careers, says Aimee Cardwell, CISO in residence at tech company Transcend and former CISO of UnitedHealth Group. ... CISOs who are too rigid with the rules do a disservice to their organizations and their professional prospects, says Cardwell. Such a situation recently came up in her organization, where one of her team members initially declined to permit a third-party application from being used by workers, pointing to a security policy barring such apps. ... CISOs who don’t have a firm grasp on all that they must secure won’t succeed in their roles. “If they don’t have visibility, if they can’t talk about the effectiveness of the controls, then they won’t have credibility and the confidence in them among leadership will erode,” Knisley says.

A CIO's Evolving Role in the Generative AI Era

The dual mandate facing CIOs today is demanding but unavoidable. They must deliver quick AI pilots that boards can take to the shareholders while also enforcing guardrails on security, ethics and cost aspects. Too much caution can make CIOs irrelevant. This balancing act requires not only technical fluency but also narrative skill. The ability to translate AI experiments into business outcomes that CEOs and boards can trust can make CIOs a force. The MIT report highlights another critical decision point: whether to build or buy. Many enterprises attempt internal builds, but externally built AI partnerships succeed twice as often. CIOs, pressured for fast results, must be pragmatic about when to build and when to partner. Gen AI does not - and never will - replace the CIO role. But it demands corrections. The CIO who once focused on alignment must now lead business transformation. Those who succeed will act less as CIOs and more as AI diplomats, bridging hype with pragmatism, connecting technological opportunities to shareholder value and balancing the boardroom's urgency with the operational reality. As AI advances, so does the CIO's role - but only if they evolve. Their reporting line to the CEO symbolizes greater trust and higher stakes. Unlike previous technology cycles, AI has brought the CIO to the forefront of transformation. 

Building an AI Team May Mean Hiring Where the Talent Is, Not Where Your Bank Is

Much of the adaptation of banking to AI approaches requires close collaboration between AI talent with people who understand how the banking processes involved need to work. This will put people closer together, literally, to facilitate both quick and in-depth but always frequent interactions to make collaboration work — paradoxically, increased automation needs more face-to-face dealings at the formative stages. However, the "where" of the space will also hinge on where AI and innovation talent can be recruited, where that talent is being bred and wants to work, and the types of offices that talent will be attracted to. ... "Banks are also recruiting for emerging specialties in responsible AI and AI governance, ensuring that their AI initiatives are ethical, compliant and risk-managed," the report says. "As ‘agentic AI’ — autonomous AI agents — and generative AI gain traction, firms will need experts in these cutting-edge fields too." ... Decisions don’t stop at the border anymore. Jesrani says that savvy banks look for pockets of talent as well. ... "Banks are contemplating their global strategies because emerging markets can provide them with talent and capabilities that they may not be able to obtain in the U.S.," says Haglund. "Or there may be things happening in those markets that they need to be a part of in order to advance their core business capabilities."

How Data Immaturity is Preventing Advanced AI

Data immaturity, in the context of AI, refers to an organisation’s underdeveloped or inadequate data practices, which limit its ability to leverage AI effectively. It encompasses issues with data quality, accessibility, governance, and infrastructure. Critical signs of data immaturity include inconsistent, incomplete, or outdated data leading to unreliable AI outcomes; data silos across departments hindering access and comprehensive analysis, as well as weak data governance caused by a lack of policies on data ownership, compliance and security, which introduces risks and restricts AI usage. ... Data immaturity also leads to a lack of trust in analysis and predictability of execution. That puts a damper on any plans to leverage AI in a more autonomous manner—whether for business or operational process automation. A recent study by Kearney found that organisations globally are expecting to increase data and analytics budgets by 22% in the next three years as AI adoption scales. Fragmented data limits the predictive accuracy and reliability of AI, which are crucial for autonomous functions where decisions are made without human intervention. As a result, organisations must get their data houses in order before they will be able to truly take advantage of AI’s potential to optimise workflows and free up valuable time for humans to focus on strategy and design, tasks for which most AI is not yet well suited.

From Reactive Tools to Intelligent Agents: Fulcrum Digital’s AI-First Transformation

To mature, LLM is just one layer. Then you require the integration layer, how you integrate it. Every customer has multiple assets in their business which have to connect with LLM layers. Every business has so many existing applications and new applications; businesses are also buying some new AI agents from the market. How do you bring new AI agents, existing old systems, and new modern systems of the business together — integrating with LLM? That is one aspect. The second aspect is every business has its own data. So LLM has to train on those datasets. Copilot and OpenAI are trained on zillions of data, but that is LLM. Industry wants SLM—small language models, private language models, and industry-orientated language models. So LLMs have to be fine-tuned according to the industry and also fine-tuned according to their data. Nowadays people come to realise that LLMs will never give you 100 per cent accurate solutions, no matter which LLM you choose. That is the phenomenon customers and everybody are now learning. The difference between us and others: many players who are new to the game deliver results with LLMs at 70–75 per cent. Because we have matured this game with multiple LLMs coexisting, and with those LLMs together maturing our Ryze platform, we are able to deliver more than 93–95 per cent accuracy. 

You Didn't Get Phished — You Onboarded the Attacker

Many organizations respond by overcorrecting: "I want my entire company to be as locked down as my most sensitive resource." It seems sensible—until the work slows to a crawl. Without nuanced controls that allow your security policies to distinguish between legitimate workflows and unnecessary exposure, simply applying rigid controls that lock everything down across the organization will grind productivity to a halt. Employees need access to do their jobs. If security policies are too restrictive, employees are either going to find workarounds or continually ask for exceptions. Over time, risk creeps in as exceptions become the norm. This collection of internal exceptions slowly pushes you back towards "the castle and moat" approach. The walls are fortified from the outside, but open on the inside. And giving employees the key to unlock everything inside so they can do their jobs means you are giving one to Jordan, too. ... A practical way to begin is by piloting ZSP on your most sensitive system for two weeks. Measure how access requests, approvals, and audits flow in practice. Quick wins here can build momentum for wider adoption, and prove that security and productivity don't have to be at odds. ... When work demands more, employees can receive it on request through time-bound, auditable workflows. Just enough access is granted just in time, then removed. By taking steps to operationalize zero standing privileges, you empower legitimate users to move quickly—without leaving persistent privileges lying around for Jordan to find.

OT Security: When Shutting Down Is Not an Option

some of the most urgent and disruptive threats today are unfolding far from the keyboard, in operational technology environments that keep factories running, energy flowing and transportation systems moving. In these sectors, digital attacks can lead to physical consequences, and defending OT environments demands specialized skills. Real-world incidents across manufacturing and critical infrastructure show how quickly operations can be disrupted when OT systems are not adequately protected. Just this week, Jaguar Land Rover disclosed that a cyberattack "severely disrupted" its automotive manufacturing operations. ... OT environments present challenges that differ sharply from traditional IT. While security is improving, OT security teams must protect legacy control systems running outdated firmware, making them difficult to patch. Operators need to prioritize uptime and safety over system changes; and IT and OT teams frequently work in silos. These conditions mean that breaches can have physical as well as digital consequences, from halting production to endangering lives. Training tailored to OT is essential to secure critical systems while maintaining operational continuity. ... An OT cybersecurity learning ecosystem is not a one-time checklist but a continuous program. The following elements help organizations choose training that meets current needs while building capacity for ongoing improvement.

Connected cars are racing ahead, but security is stuck in neutral

Connected cars are essentially digital platforms with multiple entry points for attackers. The research highlights several areas of concern. Remote access attacks can target telematics systems, wireless interfaces, or mobile apps linked to the car. Data leaks are another major issue because connected cars collect sensitive information, including location history and driving behavior, which is often stored in the cloud. Sensors present their own set of risks. Cameras, radar, lidar, and GPS can be manipulated, creating confusion for driver assistance systems. Once inside a vehicle, attackers can move deeper by exploiting the CAN bus, which connects key systems such as brakes, steering, and acceleration. ... Most drivers want information about what data is collected and where it goes, yet very few said they have received that information. Brand perception also plays a role. Many participants prefer European or Japanese brands, while some expressed distrust toward vehicles from certain countries, citing political concerns, safety issues, or perceived quality gaps. ... Manufacturers are pushing out new software-defined features, integrating apps, and rolling out over the air updates. This speed increases the number of attack paths and makes it harder for security practices and rules to keep up.

Circular strategies for data centers

Digital infrastructure is scaling rapidly, with rising AI workloads and increased compute density shaping investment decisions. Growth on that scale can generate unnecessary waste unless sustainability is integrated into planning. Circular thinking makes it possible to expand capacity without locking facilities into perpetual hardware turnover. Operators can incorporate flexibility into refresh cycles by working with vendors that design modular platforms or by adopting service-based models that build in maintenance, refurbishment, and recovery. ... Sustainable planning also involves continuous evaluation. Instead of defaulting to wholesale replacement, facilities can test whether assets still meet operational requirements through reconfiguration, upgrades, or role reassignment. This kind of iterative approach gives operators a way to match innovation with responsibility, ensuring that capacity keeps pace with demand without discarding equipment prematurely. ... The transition to circular practices is more than an environmental gesture. For data centers, it is a strategic shift in how infrastructure is procured, maintained, and retired. Extending lifecycles, redeploying equipment internally, refurbishing where possible, and ensuring secure, responsible recycling at the end of use all contribute to a more resilient operation in a resource-constrained and tightly regulated industry.

Daily Tech Digest - July 15, 2025

---

Quote for the day:

“Rarely have I seen a situation where doing less than the other guy is a good strategy.” -- Jimmy Spithill

---

CyberArk: Rise in Machine Identities Poses New Risks

The CyberArk report outlines the substantial business consequences of failing to protect machine identities, leaving organizations vulnerable to costly outages and breaches. Seventy-two percent of organizations experienced at least one certificate-related outage over the past year - a sharp increase compared to prior years. Additionally, 50% reported security incidents or breaches stemming from compromised machine identities. Companies that have experienced non-human identity security breaches include xAI, Uber, Schneider Electric, Cloudflare and BeyondTrust, among others. "Machine identities of all kinds will continue to skyrocket over the next year, bringing not only greater complexity but also increased risks," said Kurt Sand, general manager of machine identity security at CyberArk. "Cybercriminals are increasingly targeting machine identities - from API keys to code-signing certificates - to exploit vulnerabilities, compromise systems and disrupt critical infrastructure, leaving even the most advanced businesses dangerously exposed." ... Fifty percent of security leaders reported security incidents or breaches linked to compromised machine identities in the previous year. These incidents led to delays in application launches for 51% companies, customer-impacting outages for 44% and unauthorized access to sensitive systems for 43%.

What Can Businesses Do About Ethical Dilemmas Posed by AI?

Digital discrimination is a product of bias incorporated into the AI algorithms and deployed at various levels of development and deployment. The biases mainly result from the data used to train the large language models (LLMs). If the data reflects previous iniquities or underrepresents certain social groups, the algorithm has the potential to learn and perpetuate those iniquities. Biases may occasionally culminate in contextual abuse when an algorithm is used beyond the environment or audience for which it was intended or trained. Such a mismatch may result in poor predictions, misclassifications, or unfair treatment of particular groups. Lack of monitoring and transparency merely adds to the problem. In the absence of oversight, biased results are not discovered. ... Human-in-the-loop systems allow intervention in real time whenever AI acts unjustly or unexpectedly, thus minimizing potential harm and reinforcing trust. Human judgment makes choices more inclusive and socially sensitive by including cultural, emotional, or situational elements, which AI lacks. When humans remain in the loop of decision-making, accountability is shared and traceable. This removes ethical blind spots and holds users accountable for consequences.

Beyond the hype: AI disruption in India’s legal practice

The competitive dynamics are stark. When AI can complete a ten-hour task in two hours, firms face a pricing paradox: how to maintain profitability while passing efficiency gains to the clients? Traditional hourly billing models become unsustainable when the underlying time economics change dramatically. ... Effective AI integration hinges on a strong technological foundation, encompassing secure data architecture, advanced cybersecurity measures and a seamless and hassle-free interoperability between systems and already existing platforms. SAM’s centralised Harvey AI approach and CAM’s multi-tool strategy both imply significant investment in these backend capabilities. ... Merely automating existing workflows fails to leverage AI’s transformative potential. To unlock AI’s full transformative value, firms must rethink their legal processes – streamlining tasks, reallocating human resources to higher order functions and embedding AI at the core of decision-making processes and document production cycles. ... AI enables alternative service models that go beyond the billable hour. Firms that rethink on how they can price say, by offering subscription-based or outcome-driven services, and position themselves as strategic partners rather than task executors, will be best positioned to capture long-term client value in an AI-first legal economy.

‘Chronodebt’: The lose/lose situation few CIOs can escape

One needn’t be an expert in the field of technical architecture to know that basing a capability as essential as air traffic control on such obviously obsolete technology is a bad idea. Someone should lose their job over this. And yet, nobody has lost their job over this, nor should they have. That’s because the root cause of the FAA’s woes — poor chronodebt management, in case you haven’t been paying attention — is a discipline that’s rarely tracked by reliable metrics and almost-as-rarely budgeted for. Metrics first: While the discipline of IT project estimation is far from reliable, it’s good enough to be useful in estimating chronodebt’s remediation costs — in the FAA’s case what it would have to spend to fix or replace its integrations and the integration platforms on which those integrations rely. That’s good enough, with no need for precision. Those running the FAA for all these years could, that is, estimate the cost of replacing the programs used to export and update its repositories, and replacing the 3 ½” diskettes and paper strips on which they rely. But, telling you what you already know, good business decisions are based not just on estimated costs, but on benefits netted against those costs. The problem with chronodebt is that there are no clear and obvious ways to quantify the benefits to be had by reducing it.

Can System Initiative fix devops?

System Initiative turns traditional devops on its head. It translates what would normally be infrastructure configuration code into data, creating digital twins that model the infrastructure. Actions like restarting servers or running complex deployments are expressed as functions, then chained together in a dynamic, graphical UI. A living diagram of your infrastructure refreshes with your changes. Digital twins allow the system to automatically infer workflows and changes of state. “We’re modeling the world as it is,” says Jacob. For example, when you connect a Docker container to a new Amazon Elastic Container Service instance, System Initiative recognizes the relationship and updates the model accordingly. Developers can turn workflows — like deploying a container on AWS — into reusable models with just a few clicks, improving speed. The GUI-driven platform auto-generates API calls to cloud infrastructure under the hood. ... An abstraction like System Initiative could embrace this flexibility while bringing uniformity to how infrastructure is modeled and operated across clouds. The multicloud implications are especially intriguing, given the rise in adoption of multiple clouds and the scarcity of strong cross-cloud management tools. A visual model of the environment makes it easier for devops teams to collaborate based on a shared understanding, says Jacob — removing bottlenecks, speeding feedback loops, and accelerating time to value.

An exodus evolves: The new digital infrastructure market

Regulatory pressures have crystallised around concerns over reliance on a small number of US-based cloud providers. With some hyperscalers openly admitting that they cannot guarantee data stays within a jurisdiction during transfer, other types of infrastructure make it easier to maintain compliance with UK and EU regulations. This is a clear strategy to avoid future financial and reputational damage. ... 2025 is a pivotal year for digital infrastructure. Public cloud will remain an essential part of the IT landscape. But the future of data strategy lies in making informed, strategic decisions, leveraging the right mix of infrastructure solutions for specific workloads and business needs. As part of our research, we assessed the shape of this hybrid market. ... With one eye to the future, UK-based cloud providers must be positioned as a strategic advantage, offering benefits such as data sovereignty, regulatory compliance, and reduced latency. Businesses will need to situate themselves ever more precisely on the spectrum of digital infrastructure. Their location will reflect how they embrace a hybrid model that balances public cloud, private cloud, colocation and on-premise options. This approach will not only optimise performance and costs but also provide long-term resilience in an evolving digital economy.

How Trump's Cyber Cuts Dismantle Federal Information Sharing

"The budget cuts, personnel reductions and other policy changes have decreased the volume and frequency of CISA's information sharing activities in both formal and informal channels," Daniel told ISMG. While sector-specific ISACs still share information, threat sharing efforts tied to federal funding - such as the Multi-State ISAC, which supports state and local governments - "have been negatively affected," he said . One former CISA staffer who recently accepted the administration's deferred resignation offer told ISMG the agency's information-sharing efforts "were among the first to take a hit" from the administration's cuts, with many feeling pressured into silence. ... Analysts have also warned that cuts to cyber staff across federal agencies and risks to initiatives including the National Vulnerability Database and Common Vulnerabilities and Exposures program could harm cybersecurity far beyond U.S. borders. The CVE program is dealing with backlogs and a recent threat to shut down funding over a federal contracting issue. Failure of the CVE Program "would have wide impacts on vulnerability management efficiency and effectiveness globally," said John Banghart, senior director for cybersecurity services at Venable and a key architect of the Obama administration's cybersecurity policy as a former director for federal cybersecurity for the National Security Council.

Securing vehicles as they become platforms for code and data

Recently security researchers have demonstrated real-world attacks against connected cars, such as wireless brake manipulation on heavy trucks by spoofing J-bus diagnostic packets. Another very recent example is successful attacks against autonomous car LIDAR systems. While the distribution of EV and advanced cars becomes more pervasive across our society, we expect these types of attacks and methods to continue to grow in complexity. Which makes a continuous, real-time approach to securing the entire ecosystem (from charger, to car, to driver) even more so important. ... Over-the-air (OTA) update hijacking is very real and often enabled by poor security design, such as lack of encryption, improper authentication between the car and backend, and lack of integrity or checksum validation. Attack vectors that the traditional computer industry has dealt with for years are now becoming a harsh reality in the automotive sector. Luckily, many of the same approaches used to mitigate these risks in IT can also apply here ... When we look at just the automobile, we have a variety of connected systems which typically all come from different manufacturers (Android Automotive, or QNX as examples) which increases the potential for supply chain abuse. We also have devices which the driver introduces which interacts with the car’s APIs creating new entry points for attackers.

Strategizing with AI: How leaders can upgrade strategic planning with multi-agent platforms

Building resiliency and optionality into a strategic plan challenges humans’ cognitive (and financial) bandwidth. The seemingly endless array of future scenarios, coupled with our own human biases, conspires to anchor our understanding of the future in what we’ve seen in the past. Generative AI (GenAI) can help overcome this common organizational tendency for entrenched thinking, and mitigate the challenges of being human, while exploiting LLMs’ creativity as well as their ability to mirror human behavioral patterns. ... In fact, our argument reflects our own experience using a multi-agent LLM simulation platform built by the BCG Henderson Institute. We’ve used this platform to mirror actual war games and scenario planning sessions we’ve led with clients in the past. As we’ve seen firsthand, what makes an LLM multi-agent simulation so powerful is the possibility of exploiting two unique features of GenAI—its anthropomorphism, or ability to mimic human behavior, and its stochasticity, or creativity. LLMs can role-play in remarkably human-like fashion: Research by Stanford and Google published earlier this year suggests that LLMs are able to simulate individual personalities closely enough to respond to certain types of surveys with 85% accuracy as the individuals themselves.

The Network Challenges of IoT Integration

IoT interoperability and compatible security protocols are a particular challenge. Although NIST and ISO, among other organizations, have issued IoT standards, smaller IoT manufacturers don't always have the resources to follow their guidance. This becomes a network problem because companies have to retool these IoT devices before they can be used on their enterprise networks. Moreover, because many IoT gadgets are delivered with default security settings that are easy to undo, each device has to be hand-configured to ensure it meets company security standards. To avoid potential interoperability pitfalls, network staff should evaluate prospective technology before anything is purchased. ... First, to achieve high QoS, every data pipeline on the network must be analyzed -- as well as every single system, application and network device. Once assessed, each component must be hand-calibrated to run at the highest performance levels possible. This is a detailed and specialized job. Most network staff don't have trained QoS technicians on board, so they must go externally for help. Second, which areas of the business get maximum QoS, and which don't? A medical clinic, for example, requires high QoS to support a telehealth application where doctors and patients communicate. 

Daily Tech Digest - July 12, 2025

---

Quote for the day:

"If you do what you’ve always done, you’ll get what you’ve always gotten." -- Tony Robbins

---

Why the Value of CVE Mitigation Outweighs the Costs

When it comes to CVEs and continuous monitoring, meeting compliance requirements can be daunting and confusing. Compliance isn’t just achieved; rather, it is a continuous maintenance process. Compliance frameworks might require additional standards, such as Federal Information Processing Standards (FIPS), Federal Risk and Authorization Management Program (FedRAMP), Security Technical Implementation Guides (STIGs) and more that add an extra layer of complexity and time spent. The findings are clear. Telecommunications and infrastructure companies reported an average of $3 million in new revenue annually by improving their container security enough to qualify for security-sensitive contracts. Healthcare organizations averaged $7.3 million in new revenue, often driven by unlocking expansion into compliance-heavy markets. ... The industry has long championed “shifting security left,” or embedding checks earlier in the pipeline to ensure security measures are incorporated throughout the entire software development life cycle. However, as CVE fatigue worsens, many teams are realizing they need to “start left.” That means: Using hardened, minimal container images by default; Automating CVE triage and patching through reproducible builds; Investing in secure-by-default infrastructure that makes vulnerability management invisible to most developers

Generative AI: A Self-Study Roadmap

Building generative AI applications requires comfort with Python programming and basic machine learning concepts, but you don't need deep expertise in neural network architecture or advanced mathematics. Most generative AI work happens at the application layer, using APIs and frameworks rather than implementing algorithms from scratch. ... Modern generative AI development centers around foundation models accessed through APIs. This API-first approach offers several advantages: you get access to cutting-edge capabilities without managing infrastructure, you can experiment with different models quickly, and you can focus on application logic rather than model implementation. ... Generative AI applications require different API design patterns than traditional web services. Streaming responses improve user experience for long-form generation, allowing users to see content as it's generated. Async processing handles variable generation times without blocking other operations. ... While foundation models provide impressive capabilities out of the box, some applications benefit from customization to specific domains or tasks. Consider fine-tuning when you have high-quality, domain-specific data that foundation models don't handle well—specialized technical writing, industry-specific terminology, or unique output formats requiring consistent structure.

Announcing GenAI Processors: Build powerful and flexible Gemini applications

At its core, GenAI Processors treat all input and output as asynchronous streams of ProcessorParts (i.e. two-way aka bidirectional streaming). Think of it as standardized data parts (e.g., a chunk of audio, a text transcription, an image frame) flowing through your pipeline along with associated metadata. This stream-based API allows for seamless chaining and composition of different operations, from low-level data manipulation to high-level model calls. ... We anticipate a growing need for proactive LLM applications where responsiveness is critical. Even for non-streaming use cases, processing data as soon as it is available can significantly reduce latency and time to first token (TTFT), which is essential for building a good user experience. While many LLM APIs prioritize synchronous, simplified interfaces, GenAI Processors – by leveraging native Python features – offer a way for writing responsive applications without making code more complex. ... GenAI Processors is currently in its early stages, and we believe it provides a solid foundation for tackling complex workflow and orchestration challenges in AI applications. While the Google GenAI SDK is available in multiple languages, GenAI Processors currently only support Python.

Scaling the 21st-century leadership factory

Identifying priority traits is critical; just as important, CEOs and their leadership teams must engage early and often with high-potential employees and unconventional thinkers in the organization, recognizing that innovation often comes from the edges of the business. Skip-level meetings are a powerful tool for this purpose. Most famously, Apple’s Steve Jobs would gather what he deemed the 100 most influential people at the company, including young engineers, to engage directly in strategy discussions—regardless of hierarchy or seniority. ... A culture of experimentation and learning is essential for leadership development—but it must be actively pursued. “Instillation of personal initiative, aggressiveness, and risk-taking doesn’t spring forward spontaneously,” General Jim Mattis explained in his 2019 book on leadership, Call Sign Chaos. “It must be cultivated for years and inculcated, even rewarded, in an organization’s culture. If the risk-takers are punished, then you will retain in your ranks only the risk averse,” he wrote. ... There are multiple ways to streamline decision-making, including redefining decision rights to focus on a handful of owners and distinguishing between different types of decisions, as not all choices are high stakes. 

Lessons learned from Siemens’ VMware licensing dispute

Siemens threatened to sue VMware if it didn’t provide ongoing support for the software and handed over a list of the software it was using that it wanted support for. Except that the list included software that it didn’t have any licenses for, perpetual or otherwise. Broadcom-owned VMware sued, Siemens countersued, and now the companies are battling over jurisdiction. Siemens wants the case to be heard in Germany, and VMware prefers the United States. Normally, if unlicensed copies of software are discovered during an audit, the customer pays the difference and maybe an additional penalty. After all, there are always minor mistakes. The vendors try to keep these costs at least somewhat reasonable, since at some point, customers will migrate from mission-critical software if the pain is high enough. ... For large companies, it can be hard to pivot quickly. Using open-source software can help reduce the risk of unexpected license changes, and, for many major tools there are third-party service providers that can offer ongoing support. Another option is SaaS software, Ringdahl says, because it does make license management a bit easier, since there’s usually transparency both for the customer and the vendor about how much usage the product is getting.

Microsoft says regulations and environmental issues are cramping its Euro expansion

One of the things that everyone needs to consider is how datacenter development in Europe is being enabled or impeded, Walsh said. "Because we have moratoriums coming at us. We have communities that don't want us there," she claimed, referring particularly to Ireland where local opposition to bit barns has been hardening because of the amount of electricity they consume and their environmental impact. Another area of discussion at the Datacloud keynote was the commercial models for acquiring datacenter capacity, which it was felt had become unfit for the new environment where large amounts are needed quickly. "From our perspective, time to market is essential. We've done a lot of leasing in the last two years, and that is all time for market pressure," Walsh said. "I also manage land acquisition and land development, which includes permitting. So the joy of doing that is that when my permits are late, I can lease so I can actually solve my own problems, which is amazing, but the way things are going, it's going to be very difficult to continue to lease the infrastructure using co-location style funding. It's just getting too big, and it's going to get harder and harder to get up the chain, for sure," she explained. ... "European regulations and planning are very slow, and things take 18 months longer than anywhere else," she told attendees at <>Bisnow's Datacenter Investment Conference and Expo (DICE) in Ireland.

350M Cars, 1B Devices Exposed to 1-Click Bluetooth RCE

The scope of affected systems is massive. The developer, OpenSynergy, proudly boasts on its homepage that Blue SDK — and RapidLaunch SDK, which is built on top of it and therefore also possibly vulnerable — has been shipped in 350 million cars. Those cars come from companies like Mercedes-Benz, Volkswagen, and Skoda, as well as a fourth known but unnamed company. Since Ford integrated Blue SDK into its Android-based in-vehicle infotainment (IVI) systems in November, Dark Reading has reached out to determine whether it too was exposed. ... Like any Bluetooth hack, the one major hurdle in actually exploiting these vulnerabilities is physical proximity. An attacker would likely have to position themselves within around 10 meters of a target device in order to pair with it, and the device would have to comply. Because Blue SDK is merely a framework, different devices might block pairing, limit the number of pairing requests an attacker could attempt, or at least require a click to accept a pairing. This is a point of contention between the researchers and Volkswagen. ... "Usually, in modern cars, an infotainment system can be turned on without activating the ignition. For example, in the Volkswagen ID.4 and Skoda Superb, it's not necessary," he says, though the case may vary vehicle to vehicle. 

Leaders will soon be managing AI agents – these are the skills they'll need, according to experts

An AI agent is essentially just "a piece of code", says Jarah Euston, CEO and Co-Founder of AI-powered labour platform WorkWhile, which connects frontline workers to shifts. "It may not have the same understanding, empathy, awareness of the politics of your organization, of the fears or concerns or ambitions of the people around that it is serving. "So managers have to be aware that the agent is only as good as how you've trained it. I don't think we're close yet to having agents that can operate without any human oversight. "As a manager, you want to leverage the AI to make you and your team more productive, but you constantly have to be checking, iterating and training your tools to get the most out of them."  ... Technological skills are expected to become increasingly vital over the next five years, outpacing the growth of all other skill categories. Leading the way are AI and big data, followed closely by networking, cybersecurity and overall technological literacy. The so-called 'soft skills' of creative thinking and resilience, flexibility and agility are also rising in importance, along with curiosity and lifelong learning. Empathy is one skill AI agents can't learn, says Women in Tech's Moore Aoki, and she believes this will advantage women.

Common Master Data Management (MDM) Pitfalls

In addition to failing to connect MDM’s value with business outcomes, “People start with MDM by jumping in with the technology,” Cooper said. “Then, they try to fit the people, processes, and master data into their selected technology.” Moreover, in the process of prioritizing technology first, organizations take for granted that they have good data quality, data that is clean and fit for purpose. Then, during a major initiative, such as migrating to a cloud environment, they discover their data is not so clean. ... Organizations fall into the pitfalls above and others because they try to do it alone, and most have never done MDM before. Instead, “Organizations have different capabilities with MDM,” said Cooper, “and you don’t know what you don’t know.” ... Connecting the MDM program to business objectives requires talking with the stakeholders across the organization, especially divisions with direct financial risks such as sales, marketing, procurement, and supply. Cooper said readers should learn the goals of each unit and how they measure success in growing revenue, reducing cost, mitigating risk, or operating more efficiently. ... Cooper advised focusing on data quality – e.g., through reference data – rather than technology. In the figure below, a company has data about a client, Emerson Electric, as shown on the left. 

Why Cloud Native Security Is More Complex Than You Think

Enterprise security tooling can help with more than just the monitoring of these vulnerabilities though. And, often older vulnerabilities that have been patched by the software vendor will offer “fix status” advice. This is where a specific package version is shown to the developer or analyst responsible for remediating the vulnerability. When they upgrade the current package to that later version, the vulnerability alert will be resolved. To confuse things further, the applications running in containers or serverless functions also need to be checked for non-compliance. Warnings that may be presented by security tooling when these applications are checked against recognised compliance standards, frameworks or benchmarks for noncompliance are wide and varied. For example, if a serverless function has overly permissive access to another cloud service and an attacker gets access to the serverless function’s code via a vulnerability, the attack’s blast radius could exponentially increase as a result. Or, often compliance checks reveal how containers are run with inappropriate network settings. ... At a high level, these components and importantly, how they interact with each other, is why applications running in the cloud require time, effort and specialist expertise to secure them.

Daily Tech Digest - June 09, 2025

---

Quote for the day:

"Motivation gets you going and habit gets you there." -- Zig Ziglar

---

Architecting Human-AI Relationships: Governance Frameworks for Emotional AI Integration

The talent retention implications prove equally compelling, particularly as organizations compete for digitally native workforce demographics who view AI collaboration as a natural extension of professional relationships. ... Perhaps most significantly, healthy human-AI collaboration frameworks unleash innovation potential that traditional technology deployment approaches consistently fail to achieve. When teams feel psychologically safe in their AI partnerships—confident that transitions will be managed thoughtfully and that their emotional investment in digital collaborators is acknowledged and supported—they demonstrate a remarkable willingness to explore advanced AI capabilities, experiment with novel applications, and push the boundaries of what artificial intelligence can accomplish within organizational contexts. ... The ultimate result is organizational resilience that extends far beyond technical robustness. Comprehensive governance approaches that address technical performance and psychological factors create AI ecosystems that adapt gracefully to technological change, maintain continuity through system transitions, and sustain collaborative effectiveness across the inevitable evolution of artificial intelligence capabilities.

CISOs reposition their roles for business leadership

“The CISOs of the present and the future need to get out of being just technologists and build their influence muscle as well as their communication muscle,” Kapil says. They need to be able to “relay the technology and cyber messaging in words and meanings where a non-technologist actually understands why we’re doing what we’re doing.” ... “CISOs who are enablers can have the greatest impact on the business because they understand the business objectives,” LeMaire explains. “I like to say we don’t do cybersecurity for cybersecurity’s sake. … Ultimately, we do cybersecurity to contribute to the goals, missions, and objectives of the greater organization. When you’re an enabler that’s what you’re doing.” ... The BISO role emerged to bridge the gap between business objectives and cybersecurity oversight that has existed in many companies, Petrik says. “By acting as a liaison between business, technology, and cybersecurity teams, the BISO ensures that security measures are aligned with business strategies and integrated effectively,” he says. Digital transformation, emerging technologies, and rapid innovation are business mandates, and security teams add value and manage risk better when they are involved before a platform is selected or implemented, he says.

Balancing Safety and Security in Software-Defined Vehicles

Features such as Bluetooth, Wi-Fi, and cellular networks improve user convenience but create multiple attack vectors. For example, infotainment systems, because of their connectivity, are prime targets on software-defined vehicles. The recent Nissan LEAF hack revealed exactly this vulnerability, with researchers using the vehicle’s infotainment system as an entry point to access critical vehicle controls, including the steering. Not only can attackers gain access to data and location information, they can use vulnerable infotainment systems as an on-ramp to access other critical vehicle systems, like Advanced Driver Assistance Systems (ADAS), CAN-Bus, or key engine control units. ... Real-Time Operating Systems play a key role in the functionality of software-defined vehicles, as they enable precise, time-critical operations for systems like Electronic Control Units (ECUs). ECUs are primarily programmed in C and C++ due to the need for efficiency and performance in resource-constrained environments. ... Memory-based vulnerabilities, inherent to C/C++ programming, can be exploited to enable remote code execution, potentially compromising critical safety and performance systems. This creates serious cybersecurity and reliability concerns for vehicles. As RTOS suppliers manage numerous processes, any vulnerability in their codebase can be a gateway for attackers, increasing the likelihood of malicious exploits across the interconnected vehicle ecosystem.

The agile blueprint for simplifying performance management: Rethinking reviews for real impact

Understanding performance has a psychological side to it. Recognising this effect on performance frameworks, Rashmi suggested that imposter syndrome can be mitigated by making progress visible. “When you see your results in real time, you can’t keep criticising yourself.” The panellists encouraged managers to have personal discussions with their team members, which would help them build bonds. Rashmi highlighted this aspect, which can be leveraged through AI. “If AI says that there has been no potential feedback for the employee in the last month, then let the technology help the manager remind.” She also added, “Scaling up makes the quarterly reviews an exercise; hence, spontaneous quarterly check-ins are important.” Rashmi also advocated for weekly, human-centred check-ins, features that are integrated in HRStop, where it won’t be just about tracking project status, but to understand employees as people. “Treat it like a family discussion,” Rashmi recommended. “A touch of personal conversation builds deeper rapport.” Another aspect that came up in the discussion was coaching. Vimal emphasised that coaching must happen at all levels—from CXOs to interns. “It’s this cultural consistency that builds trust, retention, and performance”, he added.

Is this the perfect use case for police facial recognition?

First, as the judge noted, “fortunately the technology available prevented physical contact going further”. Availability is important here, not just in terms of the equipment being accessible; it has a specific legal element too. Where the technological means to prevent inhumane or degrading treatment are reasonably available to the police, the law in England and Wales may not just permit the use of remote biometric technology, it may even require it. I’m unaware of anyone relying on this human rights argument yet and we won’t know if these conditions would have met that threshold. ... Second, the person was on the watchlist because he was subject to a court order. This was not the public under ‘general surveillance’: a court had been satisfied on the evidence presented that an order was necessary to protect the public from sexual harm from him. He breached that order by insinuating himself into the life of a 6-year-old girl and was found alone with her. He was accurately matched with the watchlist image. The third feature is that the technology did its job. It would be easy to celebrate this as a case of ‘thank goodness nothing happened’ but that would underestimate its significance and miss the legal areas where FRT will be challenged. 

IT leaders’ top 5 barriers to AI success

Data quality issues are a real concern and an actual barrier to AI adoption, but the problem is much larger than the traditional and typical discussion about data quality in transactional or analytical environments, says John Thompson, senior vice president and principal at AI consulting firm The Hackett Group. “With gen AI, literally 100% of an organization’s data, documents, videos, policies, procedures, and more are available for active use,” Thompson says. This is a much larger issue than data quality in systems such as enterprise resource planning (ERP) or customer relationship management (CRM), he says. ... Organizations need the infrastructure in place to educate and train its employees to understand the capabilities and limitations of AI, Ally’s Muthukrishnan says. “Without the right training, adoption and utilization will not achieve the outcome you’re hoping for,” he adds. “While I believe AI is one of the largest tech transformations of our lifetime, integrating it into day-to-day processes is a huge change management undertaking.” ... “The skills gap is only going to grow,” Hackett Group’s Thompson says. “Now is the time to start. You can start with your team. Have them work on test cases. Have them work on personal projects. Have them work on passion projects. [Taking] time for everyone to take a class is just elongating the process to close the skills gap. ...”

Google’s Cloud IDP Could Replace Platform Engineering

Much of the work behind the Google Cloud IDP comes from Anna Berenberg, an engineering fellow with Google Cloud who has been with the company for 19 years. “She is the originator of a lot of these concepts overall … many of these ideas which I did not really understand the impact of until I saw it manifest itself,” said Seroter. “She had this vision that I did not even buy into three years ago. She saw a little further ahead from there, and she has built and published things. It is impressive to have such interesting engineering thought leadership, not just applied to how Google does platforms, but now turning that into how we can change … infrastructure to make it simpler. She is a pioneer of that.” In an interview with The New Stack, Berenberg said that her ideas on the IDP came to her when she looked at how this could all work using Google’s vast compute and services resources to reimagine how platform engineering could be improved. “The way it works is you have a cloud platform, and then on top of it is this thick layer of platform engineering stuff, right?” said Berenberg. “So, platform engineering teams are building a layer on top of infrastructure cloud to do an abstraction and workflows and whatever they need” to improve processes for developers. “It shrinks down because everything shifts down to the platform and now we are providing platform engineering. “

FakeCaptcha Infrastructure HelloTDS Infects Millions of Devices With Malware

The campaign’s cunning blend of social engineering and technical subterfuge has enabled threat actors to compromise systems across a vast array of regions, targeting unsuspecting users as they consume streaming media, download shared files, or even browse legitimate-appearing websites. Gendigital researchers first identified HelloTDS as an intricate Traffic Direction System (TDS) — a malicious decision engine that leverages device and network fingerprinting to select which visitors receive harmful payloads, ranging from infostealers like LummaC2 to fraudulent browser updates and tech support scams. Entry points for the menace include compromised or attacker-operated file-sharing portals, streaming sites, pornographic platforms, and even malvertising embedded in seemingly innocuous ad spots. The system’s filtering and redirection logic allows it to avoid obvious honeytraps such as virtual machines, VPNs, or known analyst environments, significantly complicating detection and takedown efforts. The scale of the campaign is staggering. Gen’s telemetry reported over 4.3 million attempted infections within just two months, with the highest impact in the United States, Brazil, India, Western Europe, and, proportionally, several Balkan and African countries.

Cutting-Edge ClickFix Tactics Snowball, Pushing Phishing Forward

ClickFix first came to light as an attack method last year when Proofpoint researchers observed compromised websites serving overlay error messages to visitors. The message claimed that a faulty browser update was causing problems, and asked the victim to open "Windows PowerShell (Admin)" (which will open a User Account Control (UAC) prompt) and then right-click to paste code that supposedly "fixed" the problem — hence the attack name. Instead of a fix, though, users were unwittingly installing malware — in that case, it was the Vidar stealer. ... "The goals of ClickFix campaigns vary depending on the attacker," says Nathaniel Jones, vice president of security and AI strategy at Darktrace. "The aim might be to infect as many systems as possible to build out a network of proxies to use later. Some attackers are trying to exfiltrate credentials or domain controller files and then sell to other threat actors for initial access. So there isn't one type of victim or one objective — the tactic is flexible and being used in different ways." ... The approach, and ClickFix in general, represents a significant innovation in the world of phishing, according to Jones, because unlike an email asking someone to click on a typosquatted link that can be easily checked, the entire attack takes place inside the browser.

Like humans, AI is forcing institutions to rethink their purpose

The institutions in place now were not designed for this moment. Most were forged in the Industrial Age and refined during the Digital Revolution. Their operating models reflect the logic of earlier cognitive regimes: stable processes, centralized expertise and the tacit assumption that human intelligence would remain preeminent. ... But the assumptions beneath these structures are under strain. AI systems now perform tasks once reserved for knowledge workers, including summarizing documents, analyzing data, writing legal briefs, performing research, creating lesson plans and teaching, coding applications and building and executing marketing campaigns. Beyond automation, a deeper disruption is underway: The people running these institutions are expected to defend their continued relevance in a world where knowledge itself is no longer as highly valued or even a uniquely human asset. ... This does not mean institutional collapse is inevitable. But it does suggest that the current paradigm of stable, slow-moving and authority-based structures may not endure. At a minimum, institutions are under intense pressure to change. If institutions are to remain relevant and play a vital role in the age of AI, they must become more adaptive, transparent and attuned to the values that cannot readily be encoded in algorithms: human dignity, ethical deliberation and long-term stewardship.