Commerce Lifts Fable Export Controls After 19-Day Halt as Anthropic Tightens AI Safeguards
Updated
Updated · POLITICO · Sep 20
Commerce Lifts Fable Export Controls After 19-Day Halt as Anthropic Tightens AI Safeguards
2 articles · Updated · POLITICO · Sep 20
Summary
Commerce restored access to Anthropic’s Fable on July 1 after 19 days offline, ending export controls imposed when the model was pulled over jailbreak concerns.
Anthropic won the reversal by sending engineers to Washington, showing updated classifiers and real-time testing data that convinced CAISI, the White House cyber office and eventually the Pentagon that safeguards had improved.
The crackdown began two days after Fable’s June 9 release, when Amazon researchers found a jailbreak; after Anthropic refused to take the model down voluntarily, Commerce ordered foreign access to Fable and Mythos suspended.
Mythos restrictions were lifted first on June 26, while Deputy Defense Secretary Stephen Feinberg held up Fable until the Defense Department completed its own review.
The episode turned a Trump administration that had promised light-touch AI policy into an active gatekeeper, leaving top AI firms operating under unpublished voluntary rules and closer White House scrutiny.
Will the threat of instant export controls force tech giants to develop future AI models entirely in secret to avoid government interference?
Could the sudden government shutdown of advanced AI inadvertently give foreign adversaries the exact advantage the US is trying to prevent?
With AI now capable of cracking encryption, are current global cybersecurity standards already obsolete against the next generation of frontier models?