Anthropic Adds New Guardrail to Regain US Government Approval for Claude Fable 5 Export
Published · updated · curated by AI Is Going Just Great
Source: wired.com ↗
"Anthropic has agreed to proactively detect and address security risks posed by the models." — Commerce Secretary Howard Lutnick
After the Trump administration imposed export controls that effectively took Anthropic's Claude Fable 5 model offline, Anthropic agreed to extend an existing safety guardrail to cover a specific behavior flagged in an Amazon research paper. The new measure blocks and reroutes to the less-capable Opus 4.8 model any requests that attempt to exploit the workaround — which, per a Luta Security analysis, involved asking Fable 5 to fix code rather than identify security issues in it, thereby sidestepping a restriction on sensitive cybersecurity capabilities. Cybersecurity experts generally don't consider this behavior alarming, but the administration's awareness of it triggered the standoff.
Commerce Secretary Howard Lutnick announced the lifting of export restrictions after the Commerce Department's Center for AI Standards and Innovation determined the model's safeguards were sufficiently robust. However, Defense Secretary Pete Hegseth has signaled there is no clear path to rescinding his February 28 order designating Anthropic a supply chain risk — meaning the company's regulatory troubles are eased but not resolved.