The “mastermind” behind Anthropic’s global removal of Mythos and Fable models: Major shareholder Amazon CEO
```
Anthropic’s AI models were halted by the U.S. government just days after public release, triggered by its major investor Amazon.
According to The Wall Street Journal, Amazon CEO Andy Jassy called U.S. Treasury Secretary Besant and other senior officials, claiming their researchers had bypassed the security guardrails of Anthropic’s flagship model Fable 5 with specific prompts, obtaining sensitive information that could be used for cyberattacks. The White House then held an emergency meeting; the Commerce Department ultimately imposed export controls on the Fable and Mythos models, banning access by overseas users. Anthropic subsequently suspended all user access to both models to comply with legal requirements.
These controls not only cast a shadow over Anthropic’s prospects as it prepares for its IPO, but also reflect the covert rivalry among tech giants, the U.S. government’s anxiety about cutting-edge AI safety, and the longstanding tensions between Anthropic and the Trump administration.
Amazon’s Late-Night Call Triggers Chain Reaction of AI Control
According to Axios, the incident began late Thursday night. Amazon called government officials and submitted a report, claiming its researchers had successfully bypassed the security restrictions of Anthropic’s Mythos model and accessed content deemed a national security threat.
The Wall Street Journal further reports that Amazon’s submission showed, with certain prompts, the Fable 5 model could uncover security vulnerabilities in at least four software programs—information normally blocked by the model’s safety mechanisms.
Andrew Morris, founder of cybersecurity firm GreyNoise Intelligence, pointed out that this information is still far from truly dangerous cyber intelligence; many public tools can also find similar vulnerabilities. Additionally, there is no evidence that Amazon researchers broke through Fable’s deeper guardrail that converts vulnerability information into executable attack code.
Anthropic responded, asserting the vulnerabilities Amazon pointed out are relatively basic problems, and that the company has robust safety mechanisms and always prioritizes security.
Failed Negotiations at Dawn, Trump Signs Export Controls
According to Axios, government officials negotiated with Anthropic for hours early Friday morning local time, demanding the voluntary removal of its latest models, but no agreement was reached.
The Wall Street Journal reports that Anthropic CEO Dario Amodei’s conversations with some officials on Friday deepened their concerns. Officials felt Anthropic displayed an unwillingness to collaborate with government security experts to address issues, reinforcing the government’s longstanding judgment of Anthropic’s lack of credibility in managing safety risks.
Ultimately, despite reservations about restricting innovation, President Trump signed and approved the control measures. White House AI advisor and venture capitalist David Sacks commented on social media that the restrictions were “reluctant,” and said “the hope now is that Anthropic fixes security issues, the export controls are lifted, and Fable is republished.”
Notably, Anthropic had repeatedly notified the U.S. government before publishing Fable and cooperated with key government AI testing agencies, adopting a phased open strategy, granting preview access priority to Amazon and other tech companies to facilitate vulnerability screening before a wider release.
Entangled Interests: Amazon Is Both Shareholder and “Whistleblower”
The complexity of this incident lies in the fact that Amazon is not only Anthropic’s main investor, but also its chip supplier, and deploys Anthropic’s best models in its own software vulnerability detection business.
The Wall Street Journal cites insiders saying Amazon CEO Jassy’s original intention was a general safety warning, but events quickly escalated into a full-scale Commerce Department ban on overseas users. An Amazon spokesperson stated that as a leading cloud provider serving large numbers of public and private clients, it is not unusual for the government to consult them about potential security risks, and that the company would not comment on the details of such discussions.
Some analysts point out that the White House’s distrust of Anthropic has longstanding roots, linked to the company's connections with progressive political donors, its public warnings about AI dangers, and its previous hiring of multiple Biden administration officials.
Valuation Pressure, User Exodus: OpenAI May Benefit
The controls happened at a sensitive time, as Anthropic is sprinting towards an IPO.
Reportedly, the company was planning to go public as early as this fall. The forced removal of its flagship model could drive users toward other platforms and exert pressure on its valuation. OpenAI is viewed as a potential beneficiary, with powerful cybersecurity models now opening up to clients and maintaining active communication with the Trump administration.
From a broader perspective, coupled with recent executive orders granting safety officials greater regulatory authority over AI models, and government discussions on potential stakes in AI companies, this incident marks a significant upgrade in the government’s regulatory approach to the AI industry.
Risk warning and disclaimerThe market has risks; investments should be made with caution. This article does not constitute personal investment advice, nor does it take into account individual users’ unique investment goals, financial status, or needs. Users should consider whether any views, opinions, or conclusions in the article meet their specific circumstances. Invest accordingly at your own risk. ```