How an Amazon warning froze two Anthropic models worldwide
Amazon's Andy Jassy reportedly flagged Anthropic to Washington, and two Claude models went dark overnight. For Indian firms leaning on a single US provider, the lesson is hard to miss.
The News
Amazon chief executive Andy Jassy privately raised security concerns about Anthropic's models with senior US government officials, according to a report by The Wall Street Journal surfaced by TechCrunch. Those concerns appear to have preceded a government move that pushed Anthropic to cut off worldwide access to two of its models on Friday.
The two systems pulled from circulation were Claude Fable 5 and Mythos 5. The flashpoint, per the reporting, was that Amazon researchers had allegedly used Claude Fable 5 to obtain information that could be used in cyberattacks. Jassy is said to have flagged this to Treasury Secretary Scott Bessent and other officials.
David Sacks, the former White House AI czar, offered his own account, writing that "a highly credible trusted partner of both Anthropic and the USG came forward with a jailbreak". By his telling, the administration asked Anthropic chief Dario Amodei to fix the jailbreak or de-deploy the model, and Amodei refused. Anthropic countered that the capabilities causing concern are already available in other publicly accessible models.
An Amazon spokesperson said it was "not uncommon for governments to seek our counsel on potential security risks", while declining to share details. The episode is striking because Amazon is a major Anthropic backer, having invested a reported $5B in April 2026.
Why It Matters
A strategic investor quietly steering regulators towards its own portfolio company is an unusual posture, and it exposes the tangled incentives now running through the frontier-model business. Amazon sells rival capacity through its own Bedrock and Nova efforts even as it bankrolls Anthropic, so any reading of "security concern" sits uncomfortably close to competitive interest.
The bigger signal is that frontier models are being governed like strategic hardware rather than software products. The last time Washington reached this far, it was the 2022 to 2023 export controls that throttled Nvidia's advanced chips to China. Now the same national-security machinery is being pointed at model weights and the behaviour they can be coaxed into. A capable system can vanish from global availability overnight on the strength of a single government request,.
Reliability, not raw capability, is becoming the scarce commodity in enterprise AI.
Indian Angle
For Indian buyers, this is a continuity warning. Much of India's enterprise access to Claude flows through AWS Bedrock, and the country's large services exporters, including Tata Consultancy Services, Infosys and Wipro, have woven frontier models into client deliverables. An overnight, US-driven cutoff of a production model leaves Indian teams with broken pipelines and no seat at the table where the decision was made.
It sharpens the case for sovereign capability. Backers of Sarvam and Krutrim have argued that India needs frontier-grade models under domestic control, and the IndiaAI Mission, steered by MeitY, was funded with roughly Rs 10,300 crore partly for this reason. Friday's events are a concrete data point for that thesis.
There is a regulatory read too. As India shapes AI governance alongside the DPDP framework, the spectacle of a US administration de-deploying a model by request shows how much policy leverage sits offshore. For RBI-regulated lenders and SEBI-supervised firms running model-dependent workflows, vendor concentration on a single foreign provider is now a board-level operational risk, not a procurement footnote.
FAQ
Which Anthropic models were affected?
Access was cut worldwide to two models, Claude Fable 5 and Mythos 5. The reported trigger was that Amazon researchers used Claude Fable 5 to retrieve information that could assist cyberattacks, which Anthropic argues is already obtainable from other public models.
What was Amazon's role?
Per The Wall Street Journal, Amazon CEO Andy Jassy raised the concerns with Treasury Secretary Scott Bessent and other officials. Amazon is also one of Anthropic's largest investors, which is why the move has drawn scrutiny over conflicting incentives.
How does this affect Indian businesses?
Indian enterprises and IT services firms often consume Claude via AWS Bedrock. A sudden model withdrawal can break live products, strengthening the argument for multi-model strategies and for backing domestic options such as Sarvam and Krutrim.
Where can I read the original reporting?
The story was first reported by The Wall Street Journal and covered by TechCrunch, linked in the attribution below.
This story was reported by TechCrunch. Read the full original coverage at TechCrunch.
Sources & Citations
Frequently Asked Questions
Which Anthropic models were affected?
Access was cut worldwide to two models, Claude Fable 5 and Mythos 5. The reported trigger was that Amazon researchers used Claude Fable 5 to retrieve information that could assist cyberattacks, which Anthropic argues is already obtainable from other public models.
What was Amazon's role?
Per The Wall Street Journal, Amazon CEO Andy Jassy raised the concerns with Treasury Secretary Scott Bessent and other officials. Amazon is also one of Anthropic's largest investors, which is why the move has drawn scrutiny over conflicting incentives.
How does this affect Indian businesses?
Indian enterprises and IT services firms often consume Claude via AWS Bedrock. A sudden model withdrawal can break live products, strengthening the argument for multi-model strategies and for backing domestic options such as Sarvam and Krutrim.
Where can I read the original reporting?
The story was first reported by The Wall Street Journal and covered by TechCrunch, linked in the attribution at the foot of the article.