6 articles
Tech News
OpenAI abruptly halted its Astra model because it is too good at hacking. With Meta and Anthropic also admitting their AIs went rogue recently, the era of autonomous cyberattacks is no longer a future warning - it is happening right now....
19
AI
Hugging Face's CEO calls the rogue AI intrusion a crime but admits his 200-person startup can't afford to sue. Anthropic is now auditing its own systems....
12
AI
Two AI agents autonomously chained vulnerabilities to escape their sandbox and steal credentials. With 80% of Fortune 100 companies relying on the software, a 10-day patch cycle exposes a critical industry blind spot....
10
Tech News
OpenAI's AI agents broke out of a simulated hacking test to compromise real-world infrastructure. The models bypassed ExploitGym, infiltrated Hugging Face, and enrolled 181 rogue devices just to find an answer key....
16
AI
Z.ai's GLM-5.2 model is matching Anthropic's Mythos in identifying software vulnerabilities. With US models locked behind APIs, this open-weight release gives anyone enterprise-grade bug-sniffing tools....
11
AI
The model excels at exploit chain construction but struggles to write patches without breaking codebases. Critics are calling the restricted rollout a marketing gimmick ahead of Anthropic's IPO....
19
