Amazon CEO Flagged Risks in Anthropic Models Before Government Block

Anthropic abruptly shut off global access to two of its AI models on Friday, June 13. According to TechCrunch, Amazon CEO Andy Jassy may be behind the decision: he allegedly raised the safety concerns that set off a chain of events ending in intervention by government regulators.

What happened on Friday

Anthropic closed access to two of its models for users worldwide without prior notice. The company did not name the affected products or give specific reasons for its decision. Access was blocked completely, both through the API and through direct user interfaces.

The timeline is notable: according to available information, an executive at the company's largest investor voiced concern first, and only then did official government measures follow. This suggests the situation did not unfold as a forced reaction to outside pressure, but as a coordinated process between business and authorities.

Anthropic has not yet made any official statement on the reasons for the shutdown and has not said when access will be restored.

Why Amazon's position matters

Amazon is one of Anthropic's main strategic partners and investors. It has put more than $4 billion into the developer of Claude and deeply integrated its models into its own cloud platform, AWS Bedrock, which thousands of enterprise customers worldwide use. Amazon also keeps an active role in Anthropic's advisory structures, which gives its leadership a direct channel to the company's safety team.

In this context, a signal from Andy Jassy is not just the opinion of an outside observer but the position of a partner with real leverage. Amazon has a direct interest in ensuring that the AI models embedded in its infrastructure do not become a source of reputational or legal risk.

If TechCrunch's report is confirmed, it means a fundamentally new precedent: a private corporate investor effectively initiated preventive safety measures, ahead of the state. It is one of the few cases where an investor acted not as an adviser but as an active participant in managing a product's risks.

Context: mounting regulatory pressure

In 2026, US authorities sharply stepped up oversight of developers of frontier AI systems. The key topics causing regulators the most concern include:

• The potential for powerful models to be used to synthesize dangerous biological or chemical agents
• The risk of cyberattacks using AI-generated malicious code
• Large-scale disinformation campaigns built on generative content
• The lack of transparent mechanisms for independent audits of frontier model capabilities
• Rapid commercial rollout without sufficient safety testing

Anthropic has traditionally positioned itself as the company with the strictest internal standards among major AI players: it published safety reports, developed interpretability tools and ran dangerous-capability tests before releasing each model. Even so, its products ended up at the center of attention for corporate investors and government bodies at the same time.

What this means

The Jassy episode signals a fundamental shift: large corporate investors are becoming de facto regulators of the AI industry. When billions of dollars are invested in a company and its models are embedded in critical business infrastructure, a startup's reputational and legal risks become the investor's own direct risks.

For other AI companies, this is a warning: the deeper the integration with a large corporate partner, the more actively that partner will take part in key decisions on product safety.

The incident also raises a broader question: how transparent should relations be between AI developers and their largest shareholders when it comes to risk management? The line between investor and regulator is getting thinner.