US fully lifts export restrictions on Anthropic's Fable 5 and Mythos 5

The US government lifted its remaining export restrictions on Anthropic's most advanced models, Claude Fable 5 and Claude Mythos 5. The models had been suspended on June 12 after researchers found a jailbreak workaround with cyberattack potential; a partial lift followed on June 26. Anthropic agreed to proactively detect security risks and alert the government to malicious activity.

Why it matters

The three-week arc — suspension, limited release, full restoration — sets a template for how Washington handles frontier-model security incidents: conditions and monitoring, not lasting bans. It removes a revenue and credibility overhang for Anthropic and its infrastructure partners; Amazon (a $13B investor with up to $20B more committed, and a $100B+ decade-long AWS commitment from Anthropic) and Broadcom (3.5 GW of compute supply from 2027) both stand to benefit from restored access.

What to watch

Whether the proactive-monitoring conditions become standard for other frontier labs, and any disclosure of what the monitoring covers.