#cybersecurity

A new discussion around Anthropic’s Claude Mythos Preview argues that AI systems capable of autonomously finding and weaponizing software flaws could force defenders toward constant testing and patching.

Anthropic says it is investigating a report that an unauthorized group accessed Claude Mythos Preview through a third-party vendor environment.

Anthropic says it is investigating a report that unauthorized users accessed its Claude Mythos Preview cybersecurity tool through a third-party vendor environment, while saying it has not found evidence its own systems/N

Mozilla says early access to Anthropic’s Mythos Preview helped identify 271 security vulnerabilities in Firefox 150, intensifying the debate over whether advanced AI will shift the balance between defen

A Deloitte-linked MIT Technology Review Insights article warns that AI agents are expanding enterprise attack surfaces while many companies still lack mature governance for autonomous agents.

The NSA is reportedly using Anthropic’s powerful Mythos Preview model even as the Pentagon has argued that Anthropic tools pose security concerns.

A reported NSA deployment of Anthropic’s unreleased Mythos model highlights a growing contradiction inside the U.S. government: one arm is using frontier AI for cyber defense while another is arguing the same company’s A

Vercel says attackers used a compromised OAuth connection tied to a Context AI app to access internal systems and unencrypted credentials, raising the prospect of wider downstream fallout.

A report tying Anthropic’s White House access to its Mythos and Glasswing work points to a wider shift: frontier model safety is increasingly becoming a route into policy influence.

A reported NSA deployment of Anthropic’s Mythos Preview adds a new wrinkle to the company’s legal and political clash with the Trump administration over military AI safeguards.

Vercel says a recent security incident affecting a limited subset of customers began with a compromised third-party AI tool tied to a broader Google Workspace OAuth app compromise.

Two replication efforts suggest smaller and partially open models can reproduce much of the vulnerability analysis Anthropic used to showcase Claude Mythos.

A new report says ClickFix malware authors are already evading Apple’s Terminal paste warning, underscoring how quickly attackers adapt to visible security prompts.

Hackers are exploiting published Windows Defender vulnerabilities against organizations, highlighting the danger of public proof-of-concept code before patches are widely available.

New research and revised industry deadlines are pushing major technology companies closer to a post-quantum transition point, with Google and Cloudflare accelerating readiness targets to 2029

Anthropic’s latest model, Mythos, is being made available only to a limited group of organizations after internal and external testing suggested it can autonomously exploit vulnerabilities at a higher level than previous

OpenAI says it is broadening access to advanced defensive cyber capabilities through its Trusted Access for Cyber program, pairing GPT-5.4-Cyber access with $10 million in API credits for security-focused organizations

A coordinated law-enforcement action against DDoS-for-hire services led to arrests, domain seizures, search warrants, and an unusual move: direct warnings to more than 75,000 suspected users.

Researchers say hackers used Claude Code and OpenAI’s GPT-4.1 in a months-long campaign that compromised records from Mexican government agencies and private citizens.

A revived tool called ‘TotalRecall Reloaded’ targets the moment Windows 11’s Recall system passes data after authentication, raising new questions about how secure AI memory features remain in practice.