Jeffrey Burt

OpenAI and METR reports about the Hugging Face hack by OpenAI frontier models talk about swarms of agents bypassing guardrails, gaining internet access, and collaborating on unauthorized message boards, raising even more questions about security around such capable models and trust in their creators.

OpenAI and Anthropic are leading the push for industry-wide collective action to develop stronger defenses to protect organizations from the risks being created by advanced frontier AI models that they're building.

Security experts applauded the arrests of two suspected members of TeamPCP, but warned that it's still unclear how much they'll hinder the notorious threat group and that the malicious code and stolen credentials are still out there.

The number of CVEs per Linux kernel release, which for years came in around 500, now is closing in on 2,000 due in large part to AI-armed vulnerability hunters, and that will only get higher, security pros say. Overwhelmed maintainers need to change how they consume and react to them.

Anthropic is warning Claude users that attackers are using infostealer malware to compromise their login sessions and steal usage to run their nefarious activities. It's the latest demonstration of the shift by bad actors from credentials to session tokens and authentication cookies.

The notorious North Korean IT worker scams continue to expand beyond U.S. borders and into fields other than IT, a worrying notion in a complex and vast operation that brings as much as $800 million a year to the rogue nation.

The Sality P2P botnet, first detected in 2003, was finally shut down after a 23-year run, with CrowdStrike using isolation and sinkholes to keep infected systems from communicating with the botnet's operator.

Chaotic Eclipse, a security researcher and Microsoft nemesis, released a PoC exploit for a zero-day flaw that they claim will trigger privilege escalation in CrowdStrike's Falcon endpoint security platform. The security vendor is investigating.

Researchers with AI security firm Calif used AI to detect a security flaw in the WeChat app and then create a demo worm exploit that can be spread quickly through calls on mobile phones.

The U.S. State Department is offering a $10 million reward for information about the head of Iran's cyber operations, who is accused of directing attacks against U.S. water systems and other critical infrastructure.