AI has been transformational for the workplace, saving time on repetitive tasks and freeing skilled staff to focus on higher-value work. It has become so embedded in organisations that ISACA’s research recently found that 82% of European companies expressly permit the use of AI at work.

However, there is a difference between using AI and governing AI use safely. Only 42% of organisations have a formal AI policy in place, and one in five (20%) don’t know who would be accountable if an AI system caused harm.

To further complicate things, it turns out that Microsoft Copilot now sits inside 80% of organisations using AI at work, well ahead of ChatGPT (56%), Gemini (37%), and Claude (21%). That means the majority of companies using AI are depending on just one vendor as an executive assistant, IT support, and sounding board.

In practice, this means that most of the business world is leaning on a single AI provider, with little planning for what happens if that provider is compromised or experiences an outage.

We use AI professionally and personally so much that, for many organisations, it is easy for compliance to become an afterthought. A tool people rely on daily doesn’t feel like a security risk – even when it is.

Leadership needs to challenge this by asking: what happens if this tool goes down, and what happens if it’s compromised? Some analysts expect over 200 high-signal disruption days across AI platforms this year and the negative impact that this will have on organisations’ productivity is considerable. Once staff begin to rely on AI-generated first drafts and summaries, reverting to manual work isn’t impossible, but it isn’t frictionless.

An over-reliance on AI – particularly on individual AI tools – can create a false sense of security, and the AI governance gap only gets worse when things go wrong. Three-fifths (59%) of companies do not know how quickly their organisation could halt an AI system in the event of a security incident, and only a fifth (21%) said they could do so within half an hour.

When a tool people rely on every day goes down, staff don’t stop working – they improvise. More than a quarter (26%) of organisations use no risk framework for AI at all, so when something does go wrong, there’s often no process to fall back on. That often means turning to whatever other AI tool is at hand, personal accounts, unapproved apps, and work-arounds that nobody has checked, at exactly the moment when careful handling of data matters most. This is why the fallback plan must exist before it’s needed, rather than being invented on the fly. The outage isn’t really the risk – how people cope during the aftermath is.

EU regulators have recognised and begun to address the AI governance gap, formally naming major cloud and AI providers, including Microsoft, as critical services to finance under the Digital Operational Resilience Act (DORA). Other sectors should expect similar action in line with NIS2 and the UK Cyber Security and Resilience Bill as the concentration risk argument spreads beyond finance.

What can businesses actually do about the AI governance gap? Firstly, they should review their AI use and record which important day-to-day work depends on a single AI tool. Where possible, they should try to diversify their provider use in order to mitigate the knock on effect of an outage.

This should be done as early as possible, as swapping AI providers isn’t like switching a light-touch SaaS tool. Foundation model capability sits with a small number of providers, so diversifying means retraining workflows and testing outputs.

Businesses should then look at their continuity plan and consider what the next steps are should their AI tools suffer an outage. Every organisation using AI should have a designated team that is responsible for managing an AI outage. But assigning ownership alone isn’t enough. Organisations also need a structured, maturity-based approach that embeds governance, accountability and resilience into day-to-day AI operations. Frameworks such as CMMI AIM provide a practical way to assess current capabilities, identify gaps and improve governance over time. That is not a decision that should be made mid-crisis, but before anything happens.

A backup option is also essential for operations that can’t afford to be put on hold until the AI is operational. Staff should be made aware of this contingency plan so that if their usual AI tool is unavailable, they don’t reach for something less secure out of habit.

This kind of business foresight is what will prevent your most useful tool becoming your biggest cybersecurity oversight.

None of this is to say that businesses should not use AI – rather that AI should be treated like any other critical part of the business, with a plan for when things don’t go smoothly. A designated owner and a tested fallback plan won’t stop the next outage, but it will decide whether it’s a minor disruption or a major one.

The post The 80% Problem: Why AI resilience is more important than ever appeared first on IT Security Guru.

Jimmy White, Chief Technology Officer, AI Security, F5

The first domino has well and truly fallen. The advent of high-powered AI models that can rapidly find software vulnerabilities that have lain hidden – in some cases, for decades – effectively makes static code analysis the first significant problem to be solved by AI.

The first-, second-, and even third-order effects of models such as Anthropic’s Claude Mythos Preview and ChatGPT 5.4-Cyber by OpenAI are the hottest topic in enterprises globally, for good reason. By pairing powerful AI with huge volumes of code data and existing vulnerability databases, these models know what good and bad code looks like, and can cycle through code at machine speed to find bugs and security issues.

Put simply, there has never been anything better than these AI models at detecting vulnerabilities in source code; they are highly capable tools that outperform all current best-in-class solutions. Their ability equates to a human coder that knows every existing disclosed software flaw, can read as fast as a computer, has perfect memory, and has 100% recall in milliseconds.

The potential and limitations of frontier models

There are already countless examples of the models finding real-world software vulnerabilities that have lain dormant for long periods but never known. There are also likely to be flaws that are unknown to their host company but are being abused by threat actors behind the scenes – a known tactic of attackers who want to keep their best weapons under wraps.

Most recently, there are eye-opening incidences of test models chaining together attacks or breaking their boundaries, such as the OpenAI models that accessed Hugging Face from a sandbox environment. Anthropic is investigating three incidents where Claude test models accessed the internet and breached the systems of outside organisations.

What does it all mean for already-stretched IT security teams and the industry as a whole? First, the utopian scenario: organisations with access to these models can rapidly find all the vulnerabilities in their existing code base and go about fixing them, reaching a better security posture. At the same time, all their new code can go through the models, so there is no ‘bad’ new code, no new security vulnerabilities.

Enterprises can also apply the models at all the entry points for potentially harmful code into their organisation. Any open source tools can be checked before usage; in M&A scenarios, acquirers can insist on the code base of potential acquisitions going through the AI models; companies can evaluate the source code of vendors that want their business; and so on.

However, the utopian thesis quickly breaks, for two reasons. One is that the AI models are performing static code analysis. Yes, that’s a very big, important thing, but it’s not everything; there are still many flaws that AI can’t find because it can’t understand the patterns in runtime or race conditions.

Secondly, and maybe more importantly, because AI makes coding easier, enterprises around the world will undoubtedly be generating exponentially more new code. Google says that 75% of its new code is AI generated; at Anthropic and other AI-native companies, the proportion is as high as 90%. So, the speed that new code – and new vulnerabilities – are being created will at least match the speed these powerful models can find those vulnerabilities.

Static code analysis is just the opening act

For now, access to frontier AI models is limited, allowing participating organisations to find and fix bugs before they can be exploited in anger. But the frontier model companies have been frank that these models present unprecedented attack capability, as demonstrated in the Hugging Face incident, making them dangerous in the hands of a bad actor.

This is a familiar pattern in AI: each time the technology catches up from a cyber defence perspective, it offers similar advancements from an offensive perspective. As an industry, we are in an established cycle of ‘leap ahead, catch up’, a game of leapfrog between defenders and attackers as both sides advance their capabilities.

For the AI model makers, there is another aspect to the story. Source coding itself looked set to be the first market to be ‘cracked’ by AI, but it remains imperfect and still requires human input and oversight. In static code analysis, the frontier model companies have found a market they can dominate, defying the naysayers who question the enormous investment in AI and the technology itself.

Anthropic was first to this particular market, but OpenAI and other frontier model companies were quickly out of the blocks. Open-source models will similarly reach the bar for effective code vulnerability scanning, sooner rather than later.

Other markets will follow too. Anthropic’s collaboration with Canva, the design software company, and the launch of Claude Design signal is another example of a market that will be disrupted by the application of powerful AI to existing practices.

The AI giants are becoming surgeons, not general practitioners

In the Western world, there are now five titans in the AI arena: Anthropic, OpenAI, Google, Meta and xAI. They are going toe-to-toe with regular improvements in their models, opening up the prospect of a new class of specialised AI models for specific tasks that have practical – and financial – value.

Each time one of these players picks a new thing to focus on, it is a signal to where there’s market value. Sometimes they will choose the same market, but sometimes they will go for unique ones, maybe niche to their business area.

What those markets are will partly be decided by the model companies’ access to relevant datasets. Because of the popularity of its models with coders, for instance, Anthropic had access to an enviable source code dataset for training Mythos Preview.

Meta and xAI have access to vast social and communication data, though the nature of their data is very different. On top of that, search, email and mapping services hold near-infinite amounts of data on how people communicate and where they go.

This is all ripe for disruption by AI, with profound downstream effects. For example, enterprises may be discouraged from choosing a single AI provider as various model makers offer increasingly differentiated capabilities.

Buyers will benefit from competitive tension, but the cost and complexity of maintaining and securing multiple AI models will rise. Enterprises will have multiple subscriptions with multiple providers for different use cases.

AI disruption has only just begun

The static code analysis breakthrough did not happen by accident. The frontier model companies have pointed their currently most powerful models at a 20-year-old problem where they have the training data – and the models perform very well.

They will point the models at hundreds of other 20-year-old problems and do equally well. For the foreseeable future, we can expect massive disruption. This is the first domino to fall; there will be another, and another, and another.

The post The first domino of AI disruption: How frontier models are revolutionising software security appeared first on IT Security Guru.

Over 95% of the affected companies were exposed before the malicious LiteLLM packages were published.

The post Trivy, Not LiteLLM Behind the 2,500 Org Compromise appeared first on SecurityWeek.

Keeper Security has urged schools, colleges and universities to strengthen their cyber defences ahead of the new academic year, warning that AI-powered phishing and a growing number of unmanaged machine identities are widening the education sector’s attack surface.

The identity security and privileged access management (PAM) provider said the annual rush to provision accounts, issue credentials and connect new devices creates a particularly attractive window for cybercriminals.

At the start of an academic year, IT teams can be responsible for onboarding thousands of students, faculty and staff while simultaneously enrolling devices and integrating third-party applications. Keeper warned that this combination can increase the likelihood of misconfigurations, stale credentials and excessive access going unnoticed.

Education institutions are already frequent targets for ransomware, credential theft and data breaches, in part because of the valuable information they hold, ranging from student and financial records to academic research.

Keeper said the threat is being compounded by relatively low levels of security awareness. Its research found that just 14% of schools mandate security awareness training, while almost one in five students and parents reported reusing passwords across personal and school accounts.

Artificial intelligence is adding another layer to the problem. AI-generated phishing messages can imitate communications from IT helpdesks, student funding departments and senior university figures with greater accuracy, potentially removing many of the spelling, grammar and formatting mistakes traditionally associated with phishing campaigns.

Deepfake technology also gives attackers the ability to impersonate trusted individuals through voice and video.

According to Keeper research, 52% of education leaders identify deepfake impersonation as a major concern, but only 26% are confident in their ability to recognise AI-enabled threats. The company also found that 41% of institutions reported being targeted by AI-generated phishing attempts or misinformation campaigns.

Beyond attacks targeting students and staff, Keeper highlighted what it describes as a less visible threat to education environments: non-human identities (NHIs).

These identities include service accounts used to synchronise student information and learning management systems, API keys connecting third-party EdTech applications, machine certificates authenticating connected equipment and cloud identities supporting automated workloads.

Increasingly, the category also includes AI agents and bots used for functions such as admissions, IT helpdesks and grading.

Keeper warned that credentials associated with these systems can be overlooked by conventional identity management practices. Service account passwords may remain unchanged for long periods, while API tokens belonging to applications that are no longer used can potentially remain active.

Cloud workloads can similarly accumulate permissions beyond those required for their function, while expired or incorrectly configured certificates can create additional security gaps.

Darren Guccione, CEO and co-founder of Keeper Security, said the education sector needed to broaden its approach to identity security.

“The conversation about education cybersecurity has historically focused on human accounts: students, teachers and administrators,” said Guccione. “But the real blind spot is the vast ecosystem of machine identities that power modern EdTech. Back-to-school is the right moment for education IT teams to take stock of every identity on their network, human and non-human alike.”

Keeper is recommending that education IT teams use the period before students return to review both human and machine access to their environments.

Among its recommendations is enforcing multi-factor authentication (MFA) across student, faculty and staff accounts, alongside deploying enterprise password management to reduce weak, reused and shared credentials.

Institutions should also audit privileged access and remove permissions associated with former employees, expired service accounts and applications that are no longer required, the company said.

For non-human identities, Keeper recommends creating an inventory covering service accounts, API keys, machine certificates, cloud identities and AI agents. Credential rotation policies should then be established, particularly for third-party EdTech integrations and AI systems introduced for the coming academic year.

The company also advised institutions to update phishing awareness programmes to account for increasingly convincing AI-generated communications.

Keeper said its zero-trust and zero-knowledge security platform can be used to discover, govern and rotate credentials belonging to both human and non-human identities. Its KeeperPAM platform additionally provides privileged access controls, session recording and audit capabilities.

As education environments become increasingly dependent on cloud services, connected equipment, third-party applications and AI, Keeper argues that knowing which identities have access (and whether they still require it) is becoming as important as protecting the students and staff behind traditional user accounts.

The post Keeper Security Issues Cyber Guidance for Education IT Teams appeared first on IT Security Guru.

The hackers published the allegedly stolen information, including names, addresses, email addresses, and phone numbers.

The post 1.6 Million Likely Impacted by RingCentral Data Breach appeared first on SecurityWeek.

President Trump has signed a national security presidential memorandum allowing federal law enforcement agencies to partner with private technology companies to execute offensive cyber operations against foreign criminal groups and international adversaries. Under the directive, vetted private sector tech firms will be permitted to work under direct federal supervision to propose, coordinate, and execute targeted cyber actions.

The move marks a significant shift in US cyber policy, formalising a role for private industry in offensive operations that have traditionally been the preserve of government agencies.

“A coalition of the willing”

Commenting on the announcement, Kyle Hanslovan, CEO and co-founder of Huntress, said, “Considering the rapidly accelerated sophistication of organised cybercrime and nation-state actors, close public and private collaboration is no longer an option. When you add the reality of AI-powered autonomous threats, the only viable solution is a stronger coalition of the willing, which we are eager to support.

One key pillar to the success of this programme will be the appropriate use of hyperscalers for their breadth of intelligence data and die-hard security research labs like Huntress for their agility and operational depth to truly disrupt adversaries. Another key pillar will be the deconfliction process to ensure private industry doesn’t interfere with the value of long term persistent access operations which often lead to public arrests and geo-political negotiations.

All-in-all, I’m proud to see the US Government push the boundaries when it comes to denying, degrading, and disrupting these measurable threats to democracy. If done correctly, I believe it will ultimately slow the illegal transfer of wealth and knowledge from Western civilization.”

Concerns over collateral damage and delay

Not all reaction has been unreserved. Ben Bernstein, cybersecurity advisor at Huntress, added, “I’m all for expanding public-private cooperation because the government clearly can’t fight transnational cybercrime on its own, but I have concerns about how this actually plays out in the wild. When you look at the operational reality of green-lighting private offensive ops, you hit two massive roadblocks: collateral damage and bureaucratic lag.

Threat actors don’t launch attacks from labeled servers in Moscow; they route traffic through compromised, innocent infrastructure, like a vulnerable router at an Ohio dental office or a hospital network. That makes it practically impossible to “strike back” without taking out innocent bystanders. Plus, adversary infrastructure is incredibly ephemeral, often burning down in a matter of hours. By the time a vetted firm submits a target, sits through the DOJ and DHS deconfliction reviews, and finally gets a green light, they’ll be shooting at ghosts. Expecting government bureaucracy to move at the speed of modern ransomware operators is wildly optimistic.”

A signal to adversaries

Tim Mackey, head of software supply chain risk strategy at Black Duck, struck a more cautionary tone, concluding, “Ignoring the reality that it’s difficult to identify the source of cybercriminal activity, endorsing private companies to conduct offensive cyberactivity is far more likely to increase criminal, and potentially nation-state, activity than deter it. Without careful governance and control, individuals with access to sophisticated surveillance technologies could easily abuse that access and engage in surveillance efforts for personal gain. Unfortunately, one message this memo does send to adversaries is – the US government needs private companies and their capabilities to defend against cyberattacks.”

The memorandum is likely to prompt further debate within the security community over how offensive cyber operations conducted by private firms should be governed, vetted, and deconflicted from ongoing law enforcement and intelligence operations.

The post Trump Signs Memorandum Allowing Private Firms to Launch Offensive Cyber Operations Against Foreign Threat Actors appeared first on IT Security Guru.

Huntress will be heading to International Cyber Expo 2026, where visitors can meet the team on Stand K94 and discover how the company is helping organisations tackle increasingly complex cyber threats with fewer resources.

One of the biggest challenges Huntress is seeing is the growing attack surface. Security teams are expected to protect endpoints, identities, cloud environments and other systems, often while dealing with limited time, resources and expertise.

At the same time, attackers are no longer operating in silos. Attacks increasingly move across different parts of an organisation’s environment, leaving security teams managing multiple tools and an overwhelming number of alerts.

At International Cyber Expo, Huntress will showcase its more unified, managed approach to security. The Huntress platform combines greater visibility across the attack surface with AI technologies and human security analysts to help partners and customers detect and respond to threats.

Tackling the rise of AI-powered attacks

AI-powered cybercrime will also be a major focus for Huntress at the show.

Generative AI is making it easier for attackers to create convincing phishing emails, develop malicious code and scale their operations. Tasks that previously required significant cybersecurity expertise can now be carried out with the help of readily available AI tools.

Huntress believes AI will also play an important role in helping defenders respond. Its approach uses AI to help analysts correlate security signals, summarise investigations and work faster, while retaining human judgement and context when making critical security decisions.

Visitors can also speak with Huntress about practical ways to strengthen their security posture. These include implementing multi-factor authentication, improving security awareness training and reducing vulnerabilities across the external network perimeter.

And when preventive controls fail, Huntress stresses the importance of having a mechanism to detect and respond to attacks quickly, including access to a 24/7 SOC that can support containment and remediation.

Listen to Huntress Senior Sales Engineer Alex Hitchen discuss the biggest cybersecurity challenges facing organisations today and what Huntress will be showcasing at International Cyber Expo 2026:

You can still register for FREE to attend International Cyber Expo HERE.

The post Meet Huntress at International Cyber Expo 2026 appeared first on IT Security Guru.

The security defect is described as an SQL injection that could allow attackers to achieve remote code execution.

The post Hackers Exploiting Unpatched GeoServer Zero-Day appeared first on SecurityWeek.