Account Lockout of OpenClaw Creator by Anthropic: Details of the Conflict with OpenAI
Peter Steinberger, creator of OpenClaw and now working at OpenAI, faced an account lockout at Anthropic. On April 10, he posted a screenshot of an email about the lockout due to "suspicious activity." The lockout lasted several hours: his X post went viral with hundreds of comments, after which access to Claude was restored.
Steinberger emphasizes that he followed Anthropic's new rules: using the API with pay-as-you-go billing, not a subscription. A week earlier, Anthropic changed its policy—subscriptions no longer cover OpenClaw and similar third-party tools. Reason: high computational load from agents that run continuous reasoning loops, repeat tasks, and integrate external services.
Technical Aspects of Load on Claude API
OpenClaw allows agents to automate complex workflows, including reasoning chains and calls to external APIs. Anthropic noted that such tools generate excessive traffic:
- Continuous chain-of-thought prompting loops.
- Automatic task retries on errors.
- Multiple parallel requests to services.
This leads to peak loads on Claude's inference servers. For developers, this means optimizing: minimize retries, cache intermediate results, and implement rate limiting on the client side.
An Anthropic engineer clarified in the comments: lockouts aren't directly tied to OpenClaw, and the company is ready to help investigate incidents. However, the timeline raises questions.
Subtext: Competition Between OpenAI and Anthropic
Steinberger links the incident to his move to OpenAI. Responding to user hints, he said: "One [OpenAI] hired me, the other [Anthropic] sent legal threats." Testing Claude in OpenClaw is driven by practical needs—the model remains the most popular among tool users, ahead of ChatGPT.
Key events in the timeline:
- Introduction of Claude Dispatch in the Cowork agent for remote control.
- Anthropic's pricing policy change for OpenClaw.
Anthropic is silent on any potential connection, but for mid/senior developers, it's a signal: when working with competing LLM platforms, factor in compliance risks and API dependencies.
Key Takeaways
- Policy Change: Anthropic subscriptions no longer cover OpenClaw—only token-based API billing.
- Ban Reason: Suspicious activity, but not directly OpenClaw; access restored after public attention.
- Technical Load: OpenClaw agents create high load due to reasoning loops and integrations.
- Competition: Author's move to OpenAI fuels suspicions of legal pressure from Anthropic.
- Compatibility: Claude leads in OpenClaw; testing continues for update stability.
— Editorial Team
No comments yet.