tech
Inside the whirlwind 24 hours that led the White House to slap export controls on Anthropic
Tense calls between Anthropic's CEO and administration officials on Friday underscore how the White House is wrestling with advanced AI models.
TL;DR
- The Trump administration imposed export controls on Anthropic's Fable and Mythos 5 AI models after a frantic 24-hour effort to get the company to voluntarily withdraw the Fable model.
- Administration officials, including Treasury Secretary Scott Bessent and White House Cyber Director Sean Cairncross, expressed concerns about security risks and the ability to bypass the model's guardrails.
- Anthropic CEO Dario Amodei defended the model's safety and guardrails, arguing that identified bypasses did not pose the same risk as a broader 'jailbreak'.
- Amazon CEO Andy Jassy raised concerns about the model's guardrails, which were then reviewed by the National Security Agency, reportedly finding 'proof' of vulnerabilities.
- The administration's decision to impose export controls was described as a 'last resort' after attempts to persuade Anthropic to address the vulnerabilities failed.