tech
White House talks with Anthropic shift to setting AI security rules
The White House and Anthropic are setting AI security rules after export controls on Fable 5, POLITICO reports, aiming to manage security risks.
TL;DR
- The White House and Anthropic are developing a framework to assess AI security flaws and guide government intervention.
- This effort follows White House export controls on Anthropic's Fable 5 and Mythos 5 AI models due to a perceived security flaw (jailbreak).
- The administration and Anthropic aim to create standardized benchmarks for evaluating future jailbreaks, including bypassed safeguards, exposed capabilities, and consequences.
- Discussions involved senior administration officials, Anthropic CEO Dario Amodei, cofounder Tom Brown, head of public policy Sarah Heck, Commerce Secretary Howard Lutnick, and National Cyber Director Sean Cairncross.
- The negotiations indicate a shift toward technical standards-setting after initial disagreements over the severity of the jailbreak.