Lakera / Check Point Software"Crash test" for LLMs in AI agents

Lakera and the UK AI Security Institute have launched 'b3', a new open source benchmark. b3 is an open source security evaluation project specifically designed to protect Large Language Models (LLM) in AI agents.