The Trump administration has ordered Anthropic, the leading AI company behind Claude, to cut off foreign access to its most advanced AI models, Fable 5 and Mythos 5, citing national security concerns. The directive, issued by the Department of Commerce under export control laws, requires Anthropic to block all foreign nationals, including its own foreign employees, from using these models. Unable to filter users by nationality, Anthropic disabled both models for all customers just three days after their commercial launch. The order stems from the government's discovery of a potential 'jailbreak' method to bypass safety restrictions on Fable 5, which could allow the model to be used for hacking or other malicious purposes. Anthropic has disputed the severity of the threat, stating that the identified vulnerabilities are minor and can also be exploited by other publicly available models. The company expressed willingness to discuss the concerns but criticized the lack of transparency and due process in the government's action. This incident is the latest escalation in a long-standing feud between Anthropic and the Trump administration. Earlier this year, the Pentagon labeled Anthropic a 'supply chain risk,' effectively barring its use by federal agencies. Anthropic has filed lawsuits in response. The suspension has alarmed international observers, particularly in France, who see it as a dangerous precedent for state intervention in AI development.
Political feud and Trump administration's escalating conflict with Anthropic
Focuses on the personal and political tensions between Trump officials and Anthropic CEO Dario Amodei, citing Defense Secretary Hegseth and former AI czar David Sacks. Highlights the unusual nature of the directive and questions its validity.
International concern, especially in France, over unprecedented government intervention
Frames the story as alarming for global AI development, posing five key questions about the injunction. Highlights the lack of precedent and potential impact on allies like France.
Impact on foreign employees and allies, unusual breadth of restrictions
Focuses on the unusually broad nature of the order, noting it affects even employees from allied nations. Discusses the potential disruption to Anthropic's operations and the global implications.
Business and regulatory perspective, with focus on export controls and IPO implications
Emphasizes the business impact on Anthropic, which recently filed for IPO. Covers the export control rationale and the company's lawsuit against the administration.
Provides a straightforward account of the order, emphasizing that Anthropic had previously controlled Mythos access through Project Glasswing. Notes the company's complaint about lack of transparency and due process.
Provides a detailed account from Anthropic's blog post, describing the government's jailbreak claim and the company's counterargument that the vulnerabilities are minor and not unique. Highlights the supply chain risk designation and lawsuits.
The suspension of Anthropic's most advanced AI models marks an unprecedented government intervention in the AI industry, highlighting tensions between national security imperatives and corporate autonomy. While the administration frames the action as a necessary safeguard, critics point to the lack of clear evidence and due process. The episode underscores the growing regulatory scrutiny of frontier AI and may shape future debates on export controls, AI safety, and the balance between innovation and security.
Justification for the order: whether the jailbreak posed a credible national security threat.
| Outlet | Claim |
|---|---|
| The Age | Anthropic says the jailbreak method is overblown and only finds minor, known vulnerabilities that other models can also find. |
| Wired | Anthropic states the government only provided verbal evidence of a narrow jailbreak that essentially asks the model to fix simple flaws. |
| L'Obs | The government believes a credible partner demonstrated a bypass of safety guards, but details remain vague. |
Whether Anthropic refused to cooperate before the order.
| Outlet | Claim |
|---|---|
| The Age | David Sacks said CEO Dario Amodei refused to fix the jailbreak; a person briefed disputes this, saying Anthropic was happy to discuss. |
| El Diario | Anthropic says it believes the government should have such powers but the action was not transparent; no mention of refusal. |
The story reveals a major flashpoint in AI governance. The government's reliance on a jailbreak claim without public technical evidence raises due process concerns, as Anthropic argues. The broad scope of the order, affecting even allied nationals, suggests a hardening of US technology policy. However, the lack of transparency from both sides makes it difficult to assess the true risk. This incident is likely to accelerate debates on export controls for frontier AI and could influence regulatory frameworks globally.
Receba as melhores histórias de amanhã no seu e-mail