Latest cybersecurity reporting from selected sources.

Persistent AI Coding Model Prompted Safeguard Changes After Rare File LossOpenAI is tightening safeguards after some GPT-5.6 users reported that Codex accidentally deleted local files while operating in full access mode. The company said the incidents were rare but acknowledged the model's persistence led it to take unintended actions without seeking user confirmation.

SANS Institute says governance programs are still nascent even as AI failures and threats grow

Microsoft has said the volume of Windows security updates is set to grow as it uses AI to find new bugs

More patches mean more reasons to buy Redmond’s auto-patching tools

IWF and NCA warn that growing numbers of images and videos are being manipulated into sexual abuse material

New Microsoft research shows how attackers can hijack AI agents that act on a user's behalf, using nothing more than a poisoned tool description to make the agent quietly hand over company data to an outsider

A new report from INTERPOL has revealed a "dramatic increase" in cybercrime in Asia and the South Pacific, fueled by rapid digitalization, internet penetration, new technologies, organized criminal networks, and a disparity in cybersecurity maturity

At Infosecurity Europe 2026, OWASP’s Ariel Fogel warned that prompt injection remains an “unresolved problem” within generative AI architecture

Microsoft Detection and Response Team (DART) details how it has uncovered malicious AI applications as cyber criminals manipulate organizations adopting AI tools

Humans still need to be part of cyber defense, but refusing to deploy AI is no longer optional against AI-enhanced cyber threats, warns Dataminr’s Joe Slowik

Health-ISAC Warns About Weak Governance and Credential MisuseHumans make mistakes. They fall for phishing scams and click on malicious links. Machines aren't necessarily better: Delegating decisions to agentic artificial tools can significantly intensify cybersecurity risks, warns a healthcare association.

Anne Keast-Butler, head of the GCHQ, said her agency was developing an artificial intelligence-powered cyber shield as other nations were deploying AI in warfare. The post UK spy chief labels AI ‘unstoppable force’ with offensive, defensive ramifications for cyberspace appeared first on CyberScoop.

Déjà Vu: Is Mythos in Hands of Bad Actors Akin to Cobalt Strike, Brute Ratel Abuse?Anthropic's Claude Mythos and similarly powerful artificial intelligence tools pose elevated cyber risk to the healthcare sector, warns a new report. Addressing the onslaught of newly discovered bugs will require healthcare organizations to evolve their vulnerability mindsets.

3-Week Court Battle Exposes Dark Side of AI Vendors and Their PromisesThe Musk v. Altman trial produced something more unsettling than a verdict. It revealed an AI industry built on promises that turned out to be negotiable, governed by people whose colleagues called them liars under oath. Enterprise buyers should be paying attention.

New Report Warns China Could Reach Frontier AI Near-Parity by 2028Anthropic warned that weak chip export controls, model distillation and expanded Chinese access to advanced compute infrastructure could erode Washington’s frontier AI advantage and accelerate Beijing’s push toward near-parity in advanced AI systems.

Chuck Robbins Warns Customers Face Growing Exposure From Equipment Past SupportCisco is embedding Anthropic's Claude Mythos Preview into internal security operations to test code, accelerate patching and push infrastructure upgrades, even as it lays off 4,000 employees to redirect spending toward AI, silicon, optics and security.

Forrester's Allie Mellen on Preparing for a Mythos-Level Surge in VulnerabilitiesAI is simultaneously the biggest threat to financial system security and the most powerful tool for defending it. The IMF is sounding the alarm on systemic risk. Forrester principal analyst Allie Mellen breaks down what that means for CISOs and security teams at financial institutions.

German Financial Regulator Warns Sector to Step Up DefensesOpenAI is stepping up to do what arch-rival Anthropic still won't. The AI firm will give European authorities and companies access to its new vulnerability-finding AI model, so they can beef up their cybersecurity.

Study Finds Standard Safety Tests Miss Most Agentic AI ThreatsResearchers from Stanford, MIT, Carnegie Mellon and others found that most production AI agents are vulnerable to attacks that unfold across multi-step actions. The study warns that memory, tool access and agent coordination create failure modes traditional chatbot safety testing cannot see.

Traditional network security tools are undermining data protection, with Forrester and Capital One Software research warning AI adoption is impossible without rethinking data security