Trump Administration's Frontier AI Export Controls Spread Across Industry
AI 안전 | Sat Jun 27 2026 00:00:00 GMT+0000 (Coordinated Universal Time) | 4 sources
Following Anthropic Mythos, OpenAI's GPT 5.6 has also been caught in government approval procedures, spreading regulatory risk throughout the U.S. AI industry.
Analysis
[Trump Administration] issued export control orders targeting Anthropic Mythos and Fable models [1]
- Export control order announced on June 12
- Demanded complete blocking of foreign access
- Applies to all non-U.S. citizens including Anthropic employees
[Anthropic] took Mythos-class models offline and continued negotiations in Washington [1]
- High-intensity negotiations ongoing for 14 days
- Executives dispatched to Washington DC
- Timeline for re-release unclear
[U.S. Department of Commerce] took emergency action after Fable 5 jailbreak vulnerability report [1]
- No issues raised during pre-release testing
- Andy Jassy reported guardrail bypass method
- Confirmed prompts bypassing security code review refusals
[U.S. Government] permitted limited release of Anthropic Mythos to 'trusted partners' [2]
- Allowed restricted access for select U.S. companies
- Partial relaxation rather than full lifting
[OpenAI] released GPT 5.6 in limited preview subject to customer-by-customer government approval [3]
- Customer-by-customer government approval process
- Altman expects general release within weeks
- Same regulatory structure as Anthropic Mythos
[Anthropic] disclosed its internal philosophy that leading frontier AI ensures safety [4]
- Valued at approximately $1 trillion
- Secured government customers including the U.S. military
- Dario Amodei: pursue safety after securing competitiveness