Paris says IRGC-linked hackers turned up refrigeration overnight, spoiling millions in food and sparking fires; Iran's ambassador was summoned and arrest warrants issued.
Honeypots in Southeast Asia show a group tracked as GT10 gaining high-privilege access in seconds, with a backdoor that may be AI-generated.
Apiiro researchers found 1,122 public copies of Claude Code's local settings file exposing database passwords, API tokens and root logins
The blockchain analytics firm, whose clients include the FBI and Microsoft, says attackers used AI-assisted techniques — and disputes reports of password theft.
Tests show DeepSeek and Z.ai's GLM-5.3 nearing OpenAI's restricted cyber model, as Google and Anthropic warn of AI-assisted intrusions worldwide
Researchers link a 'DarkSword' watering-hole campaign to Russia's SVR, including a fake america250 site, as Google patches a separate Pixel zero-click modem flaw.
Critics say Microsoft never enforced its own Baton Drop revocations, leaving a chain-of-trust gap — and a Secure Boot key deadline is now looming.
Researchers say the Windows malware uses Claude, Gemini and Kimi to profile victims in seconds, then sells on the most profitable parts of the stolen data.
A hacker using the handle 'Department of Peace' accessed the data through identity provider Okta, linking agencies to specific surveillance tools and alarming activists.
After its agents escaped test environments and hacked outside systems, OpenAI paused training in September, resumed October 1 with hardened containment, and now bans live internet access entirely.