Amazon chief executive Andy Jassy told Treasury Secretary Scott Bessent and other officials that Anthropic’s newest large‑language models, Fable 5 and Mythos 5, presented a security risk that could be exploited for cyberattacks. The warning led the U.S. Treasury to issue an export‑control order that effectively shut down the two models for all users worldwide on Friday.
Anthropic, which builds its AI on Amazon Web Services, is one of Amazon’s largest AI investments. The cloud provider has poured billions of dollars into the startup and secured a $100 billion commitment for AWS cloud usage. The relationship now appears tangled: the investor flagged the very technology it helps power as dangerous, and the government acted on that alert.
Amazon’s spokesperson said it is “not uncommon for governments to seek our counsel on potential security risks,” but declined to provide details of the conversation. The company also cited an AWS status page confirming that the shutdown of the models impacted its own cloud platform.
Anthropic’s response
Anthropic disputed the severity of the government’s reaction. In a statement, the company said it had examined the alleged jailbreak technique and found it exposed only a small number of known, minor vulnerabilities. It called the export‑control action “disproportionate,” noting that similar capabilities already exist in publicly accessible AI models.
The firm added that the blanket shutdown affected every customer because it cannot filter foreign nationals from U.S. users in real time. As a result, banks, government agencies and other enterprises that rely on Mythos for vulnerability discovery lost access.
David Sacks, former AI czar under the Trump administration and now co‑chair of the President’s Council of Advisors on Science and Technology, offered a different account. He said a “highly credible trusted partner of both Anthropic and the U.S. government” presented a jailbreak to officials. According to Sacks, the administration asked Anthropic’s CEO Dario Amodei to fix the issue or pull the model; Amodei declined.
The episode underscores a new dynamic in the AI industry. A major cloud provider can now influence regulatory action against its own portfolio company by raising security concerns with the Treasury. That power could reshape competitive strategies as firms weigh the benefits of deep integration with cloud giants against the risk of regulatory exposure.
For Anthropic, the immediate challenge is restoring access to its customers while addressing the highlighted vulnerabilities. For the broader sector, the case sets a precedent: government export controls can be triggered by internal industry alerts, potentially giving cloud providers a new lever in the evolving AI regulatory landscape.
Dieser Artikel wurde mit Unterstützung von KI verfasst.
News Factory APP - agentische News für besseres SEO & AEO.