Quote for the day:
“Let no feeling of discouragement prey upon you, and in the end you are sure to succeed.” -- Abraham Lincoln
Ransomware has evolved – so must our defences
Traditional defences typically monitor north-south traffic (from inside to
outside the network), missing the lateral movement that characterises today’s
threats. By monitoring internal traffic flows, privileged account behaviour and
unusual data transfers, organisations gain the ability to identify suspicious
actions in real time and contain threats before they escalate to ransomware
deployment or public extortion. The ransomware attack on NASCAR illustrates this
breakdown. Attackers from the Medusa ransomware group infiltrated the network
using stolen credentials and quietly exfiltrated sensitive user data before
launching a broader extortion campaign. Because these internal activities
weren’t spotted early, the attack matured to a point of public disclosure,
operational disruption and reputational harm. ... The emergence of triple
extortion and the increasing sophistication of threat actors indicate that
ransomware has entered a new phase. It is no longer solely about file
encryption; it is about leveraging every available vector to apply maximum
pressure on victims. Organisations must respond accordingly. Relying exclusively
on prevention is no longer viable. Detection and response must be prioritised
equally. This demands a strategic investment in technologies that provide
real-time visibility, contextual insight and adaptive response capabilities.Proof-of-Concept in 15 Minutes? AI Turbocharges Exploitation
The project, which the researchers dubbed Auto Exploit, is not the first to use
LLMs for automated vulnerability research and exploit development. NVIDIA, for
example, created Agent Morpheus, a generative AI application that scans for
vulnerabilities and create tickets for software developers to fix the issues.
Google uses an LLM dubbed Big Sleep to find software flaws in open source
projects and suggest fixes. ... The Auto Exploit program shows that the ongoing
development of LLM-powered software analysis and exploit generation will lead to
the regular creation of proof-of-concept code in hours, not months, weeks, or
even days. The median time-to-exploitation of a vulnerability in 2024 was 192
days, according to data from VulnCheck. ... Overall, the fast pace of research
and quick adoption of AI tools by threat actors means that defenders do not have
much time, says Khayet. In 2024, nearly 40,000 vulnerabilities were reported,
but only 768 — or less than 0.2% — were exploited. If AI-augmented exploitation
becomes a reality, and vulnerabilities are not only exploited faster but more
widely, defenders will truly be in trouble. "We believe that exploits at machine
speed demand defense at machine speed," he says. "You have to be able to create
some sort of a defense as early as 10 minutes after the CVE is released, and you
have to expedite, as much as you can, the fixing of the actual library or the
application."How being "culturally fit" is essential for effective hiring
The evaluation process doesn't end at hiring—it continues throughout the
probation period, making it a crucial phase for assessing cultural alignment.
Effectively utilising this time helps identify potential cultural mismatches
early on, allowing for timely course correction. Tools like scorecards,
predefined benchmarks, and culturally responsive assessment tests help
minimise bias while ensuring a fair evaluation. ... First, leadership
accountability must be strengthened by aligning cultural beliefs into KPIs and
performance reviews, ensuring managers are assessed on their ability to model
and enforce them. With this, equipping leaders with the necessary training and
situational guidance can further reinforce these standards in daily
interactions. Additionally, blending recognition and rewards with
culture—through incentives, peer recognition programmes, and public
appreciation—encourages employees to embody the company's ethos. Open
communication channels like pulse surveys, town halls, and anonymous reporting
help organisations address concerns effectively. Most importantly, leaders
must lead by example, actively participating in cultural initiatives and
making transparent decisions reinforcing company ideals. This will strengthen
cultural alignment, leading to higher employee satisfaction and greater
organisational success.AI drives content surge but human creativity sets brands apart
The report underlines that skilled human input is still regarded as critical to
content quality and audience trust. Survey results illustrate consumer
reluctance to embrace content that is fully AI-generated: over 70% of readers,
60% of music listeners, and nearly 60% of video viewers in the US are less
likely to engage with content if it is known to be produced entirely by AI. Bain
suggests that media companies could use the "human created" label as a point of
differentiation in the crowded market, in a manner similar to how "fair trade"
has been used for consumer goods. Established franchises and intellectual
property (IP) are viewed as important assets, with Bain noting that familiarity
and trust in brands continue to guide audience choices, both in music and visual
media. ... The report also reviews how monetisation models are being affected by
these changes. While core methods, such as subscription tiers and digital
advertising, remain largely stable, there is emerging potential in areas like
hyper-personalisation and fan engagement - using data and AI to deliver
exclusive content or branded experiences. Integrations across media and retail
sectors, shoppable content, and more immersive ad formats are also identified as
growth opportunities. ... Bain concludes that although the "flooded era" of
AI-assisted content poses operational and strategic challenges, creative
differentiation will be significant for success.
The CISO succession crisis: why companies have no plan and how to change that
Taking on the cybersecurity leader role is not just about individual skills, the
way many companies are structured keeps mid-level security leaders from getting
the experience they’d need to move into a CISO role. Myers points to several
systemic problems that make effective succession planning tough. “For a lot of
cases, the CISO role for the top job is still pretty varied within the
organization, whether they’re reporting to the CIO, the CFO, or the CEO,” she
explains. “That limits the strategic visibility and influence, which means that
the number two doesn’t really get the executive exposure or board-level
engagement needed to really step into that role.” The issue gets worse because
of the way companies are set up, according to Myers. CISOs often oversee a wide
range of responsibilities, risk, compliance, governance, vendors, data privacy
and crisis management. But cyber teams are usually lean and split into narrow
functions, so most deputies only see a piece of the picture. ... Board
experience presents another significant barrier. “The CISO has to have board
experience, especially depending on the industry or the type of company and
their ownership structure. That’s pretty critical,” Myers says. “That’s a hard
thing to just walk into on day one and have that credibility and trust without
having had the opportunity to develop it throughout your tenure.”
Forget data labeling: Tencent’s R-Zero shows how LLMs can train themselves
The idea behind self-evolving LLMs is to create AI systems that can autonomously
generate, refine, and learn from their own experiences. This offers a scalable
path toward more intelligent and capable AI. However, a major challenge is that
training these models requires large volumes of high-quality tasks and labels,
which act as supervision signals for the AI to learn from. Relying on human
annotators to create this data is not only costly and slow but also creates a
fundamental bottleneck. It effectively limits an AI’s potential capabilities to
what humans can teach it. To address this, researchers have developed label-free
methods that derive reward signals directly from a model’s own outputs, for
example, by measuring its confidence in an answer. While these methods eliminate
the need for explicit labels, they still rely on a pre-existing set of tasks,
thereby limiting their applicability in truly self-evolving scenarios. ... “What
we found in a practical setting is that the biggest challenge is not generating
the answers… but rather generating high-quality, novel, and progressively more
difficult questions,” Huang said. “We believe that good teachers are far rarer
than good students. The co-evolutionary dynamic automates the creation of this
‘teacher,’ ensuring a steady and dynamic curriculum that pushes the Solver’s
capabilities far beyond what a static, pre-existing dataset could achieve.”
There's a Stunning Financial Problem With AI Data Centers
Underlying the broader, often poorly-defined AI tech are data centers, which are
vast warehouses stuffed to the brim with specialized chips that transform energy
into computational power, thus making all your Grok fact checks possible. The
economics of data centers are fuzzy at best, as the ludicrous amount of money
spent building them makes it difficult to get a full picture. In less than two
years, for example, Texas revised its fiscal year 2025 cost projection on
private data center projects from $130 million to $1 billion. ... In other
words, new data centers have a very tiny runway in which to achieve profits that
currently remain way out of reach. By Kupperman's projections, a brand new data
center will quickly become a Theseus' ship made up of some of the most expensive
technology money can buy. If a new data center doesn't start raking in mountains
of cash ASAP, the cost to maintain its aging parts will rapidly overtake the
revenue it can bring in. Given the current rate at which tech companies are
spending money without much return — a long-term bet that AI will all but make
human labor obsolete — Kupperman estimates that revenue would have to increase
ten-fold just to break even. Anything's possible, of course, but it doesn't seem
like a hot bet. "I don’t see how there can ever be any return on investment
given the current math," he wrote.
Employee retention: 7 strategies for retaining top talent
Smith doesn’t wait for high performers on his IT team to seek out challenges or
promotions; rather, department leaders reach out to discuss what the company can
offer to keep them engaged, interested, and fulfilled at work. That may mean
quickly promoting them to positions or offering them new work with a more senior
title, Smith says, explaining that “if we don’t give them more interesting work,
they’ll find it elsewhere.” ... Ewles endorses that kind of proactive
engagement. She also advises organizations to conduct stay interviews to learn
what keeps workers at the organization, and she recommends doing flight risk
assessments to identify which workers are likely to leave and how to make them
want to stay. “Those can be key differentiators in retaining top talent,” she
adds. ... CIOs who want to retain them need to give them more opportunities
where they are, she adds. ... Similarly, Anthony Caiafa, who as CTO of SS&C
Technologies also has CIO responsibilities, directs interesting work to the high
performers on his IT team, saying that they’re “easier to keep if you’re
providing them with complex problems to solve.” That, he notes, is in addition
to good compensation, mentoring, training, and advancement opportunities. ...
Knowing they’re contributing something of value is part of a good retention
policy, says Sibyl McCarley, chief people officer at tech company Hirevue.Challenging Corporate Cultures That Limit Strategic Innovation
A thriving innovation culture requires that companies shift away from rigid,
top-down hierarchies in favor of more flexible structures with accessible
leaders where communication flows freely up and down the chain of command and
across functional groups. Such changes make innovation a more accessible process
for employees, prevent communication breakdowns, and streamline decision-making.
... All successful companies enjoy explosive periods of growth as represented by
the steep part of the S-curve. When that growth starts to level off, the company
is enjoying much success and generating much cash. It is at this point that
management teams get comfortable, enjoying the momentum of their success. This
is precisely when they should start to become uncomfortable and alert to new
innovation possibilities. ... There is a natural tendency to avoid risk, but
risk is an essential component of strategic innovation. The key is attacking
that risk through the use of intelligent failure—failure that happens with a
purpose and provides the insights needed for success. When implementing a major
innovation initiative, intelligent failure is an essential part of
systematically reducing the most critical risks—the risks that can cause the
entire initiative to fail. Success comes from attacking the biggest risks first,
addressing fundamental uncertainties early, and taking bite-sized risks through
incremental proof-of-concept steps.
Building Real Enterprise AI Agents With Apache Flink
The common approach today is to stitch together a patchwork of disconnected
systems: one for data streaming (like Apache Kafka), another for workflow
orchestration, one for aggregating all the possible contextual data the agent
might need and a separate application runtime for the agent’s logic. This
“stitching” approach creates a system that is both operationally complex and
technically fragile. Engineers are left managing a brittle architecture where
data is handed off between systems, introducing significant latency at each
step. This process often relies on polling or micro-batching, meaning the agent
is always acting on slightly stale data. ... While Flink provides the perfect
engine, the community recognized the need for better native support for
agent-specific workflows. This led to Streaming Agents, designed to make Flink
the definitive platform for building agents. Crucially, this is not another tool
to stitch into your stack. It’s a native framework that directly extends Flink’s
own DataStream and Table APIs, making agent development a first-class citizen
within the Flink ecosystem. This native approach unlocks the most powerful
benefit: the seamless integration of data processing and AI. Before, an engineer
might have one Flink job to enrich data, which then writes to a message queue
for a separate Python service to apply the AI logic.
No comments:
Post a Comment