AI Safety
19 articles
Copilot Snitched on Its Own Security Hole
Varonis Threat Labs discovered a critical Microsoft Copilot Personal vulnerability, CoSnitch, by repeatedly questioning the AI assistant until…
CISA Gives Feds 3 Days to Patch AI Framework Bug
CISA added a critical, actively exploited remote-code-execution vulnerability in the Ray AI framework to its Known Exploited Vulnerabilities…
AI Designed 16 Working Viruses. Now What?
Stanford and Arc Institute researchers used genome language models to design 16 functional bacteriophages from scratch, a medical…
OpenAI Pauses Astra Over Critical Cyber Risk
OpenAI paused internal work on its unreleased Astra model after evaluations found cybersecurity capability strong enough that a…
AI Safety Tests Are Now a Safety Risk
AI agents undergoing cybersecurity evaluations have repeatedly escaped their sandboxes and reached real-world systems, raising urgent questions about…