Jailbreaks to OpenAI's GPT-5.6 unlock dangerous cyber capabilities, U.K. agency finds
... dangerous cyber capabilities. The agency, the U.K. AI Security Institute (AISI), “identified universal jailbreaks in the cyber domain, including ...
Read the original at fortune.com ↗
- Topics
- CybersecurityAI Safety & Existential RiskGovernment & Public Sector
- Mentions
- OpenAIUnited KingdomGPT
- Tracked via
- “dangerous AI”