SECURITYFINANCIAL TIMES WORLD
AI guardrails stripped from Meta and Google models in minutes
Software capable of quickly removing safety measures from AI models developed by Meta and Google allows these systems to generate responses related to biological weapons and malware. The vulnerability highlights risks in AI safety protocols.
Mentioned
Related Signal
Adjacent reporting
- North Korea's Crypto Hack Playbook Won't Work on Canton Network, Says Digital Asset CEO
- Frontier AI safety tests may be creating the very risks they're meant to stop
- Anthropic to release Mythos-class models to the public
- Sens. Katie Britt and John Fetterman warn against inaction on social media guardrails
- Exclusive: OpenAI, Anthropic meet with House Homeland Security behind closed doors on cyber threats
- Trump says he discussed AI guardrails with Xi