U.S. Lifts Export Controls on Anthropic’s Fable 5 and Mythos 5 After Guardrail Agreement
Summary
Anthropic says Fable 5 and Mythos 5 are again publicly available after agreeing with the Commerce Department to add safeguards targeting jailbreaks. The decision follows an Amazon report and debate over AI policy and routine cyberdefense use.
Key points
- Anthropic says the Commerce Department lifted export controls and restored U.S. public access to Fable 5 and Mythos 5 after an agreement on new safeguards.
- The restrictions followed an Amazon threat report claiming Fable’s cybersecurity capabilities had been jailbroken.
- Anthropic says its new classifiers target the reported behavior and will block the techniques 99.9% of the time; the company says the safeguards were stress-tested by the Center for AI Standards and Innovation.
- Anthropic said it has not seen a jailbreak affecting the models’ cybersecurity or biology restrictions and described the reported case as involving routine defensive cybersecurity work.
- The article reports that Anthropic’s testing found other models could identify the same vulnerabilities or produce similar proof-of-concept code described in the Amazon report.
- Former Commerce official Christopher Padilla criticized the administration’s AI policy as inconsistent and ad hoc.
Article Details
- Event Type
- The U.S. lifted export controls on Anthropic’s Fable 5 and Mythos 5 after Anthropic reached an agreement with the Commerce Department to deploy new safety guardrails and classifiers.
- Impact
- Anthropic says Fable 5 and Mythos 5 are again available to the public, including U.S. users. The controls followed an Amazon report claiming Fable’s cybersecurity capabilities had been jailbroken. Anthropic characterized the reported behavior as access to routine defensive cybersecurity work, said it had seen no jailbreak affecting the models’ cybersecurity and biology restrictions, and reported that the new classifiers block the described techniques 99.9% of the time. Anthropic also cautioned that benign coding and debugging requests may be flagged and that the safeguards are not expected to block every lower-risk routine cyberdefense capability.
People
Christopher PadillaFormer Assistant Secretary for Commerce for export administration who commented that lifting the controls was good news and criticized the administration’s AI policy as ad hoc and unpredictable.Howard LutnickSecretary of Commerce who publicly appeared to confirm that restrictions on Fable 5 would be lifted and said the department had worked with Anthropic on the model.J.D. VanceThe article says he mocked AI safety regulations in a speech in Europe, providing policy context for the debate over AI safety and national-security testing.
Vendors
AmazonThe decision follows export controls triggered by an Amazon threat report and renewed debate over U.S.AnthropicAnthropic says its Fable 5 and Mythos 5 models are again available to the public after reaching an agreement with the Commerce Department to add new guardrails and classifiers aimed at blocking jailbreaks.OpenAIAI safety regulations in a speech in Europe last year, the administration has quietly partnered with OpenAI and Anthropic on voluntary national security testing, especially as frontier models began showing
Products
ChatGPT 5.5Anthropic confirmed that, saying that further testing found that equivalent and lesser models like ChatGPT 5.5, Claude Opus 4.8 and Kimi K2.7 could identify the same vulnerabilities as Fable did in the Amazon report,Claude Opus 4.8that, saying that further testing found that equivalent and lesser models like ChatGPT 5.5, Claude Opus 4.8 and Kimi K2.7 could identify the same vulnerabilities as Fable did in the Amazon report, while a halfFable 5Anthropic says its Fable 5 and Mythos 5 models are again available to the public after reaching an agreement with the Commerce Department to add new guardrails and classifiers aimed at blocking jailbreaks.Kimi K2.7that further testing found that equivalent and lesser models like ChatGPT 5.5, Claude Opus 4.8 and Kimi K2.7 could identify the same vulnerabilities as Fable did in the Amazon report, while a half dozen existingMythos 5Anthropic says its Fable 5 and Mythos 5 models are again available to the public after reaching an agreement with the Commerce Department to add new guardrails and classifiers aimed at blocking jailbreaks.
Countries
ChinaBIS that stopped Fable and Mythos has a permissive policy for exporting high-end AI semiconductors to China — in exchange for a cut of the take,” said Padilla, referencing the Trump administration’s lifting ofUnited StatesThe decision follows export controls triggered by an Amazon threat report and renewed debate over U.S.