درباره این اپیزود
Anjon Roy interviews Edward Yee, Chief Growth Officer at FAR AI, about the nonprofit’s new AI Security Leaderboard, a public ranking that measures how costly it is to find universal jailbreaks against frontier AI models providers such as Gemini, Grok, Open AI's GPT and Anthropic's Claude models, with many more to come. Yee explains why traditional evals fall short, and why a dollar-based “cost to jailbreak” metric provides the public a more actionable gauge of real-world safety. The conversation covers FAR AI’s collaborative yet independent red teaming, why CBRNE and cyber risks were prioritized, and how uneven safeguards across models create systemic vulnerabilities that bad actors can exploit.
Show Links
AI Security Leaderboard: https://leaderboard.far.ai/
FAR.AI Website: https://www.far.ai/
Edward’s LinkedIn page: https://www.linkedin.com/in/edward-yee/
Sponsor Page
Contact Show
