Anthropic Temporarily Bans OpenClaw Creator From Accessing Claude

Add us on Google

OpenAI rival Anthropic briefly suspended the account of Peter Steinberger, the creator of the popular open-source AI agent framework OpenClaw, citing “suspicious activity.”

The temporary ban occurred early on April 10, 2026, and lasted only a few hours. Steinberger shared a screenshot of the suspension notice on X, noting that it would make it harder to maintain OpenClaw’s compatibility with Anthropic’s Claude models going forward. By later that morning, after the post gained attention, his access was restored.

The incident comes amid growing tension between Anthropic and users of third-party tools like OpenClaw. Last week, Anthropic updated its policy to stop allowing Claude Pro and Max subscribers to use their flat-rate subscription limits with third-party “harnesses” or agents, including OpenClaw. Instead, users must now pay for usage through a separate pay-as-you-go API option, which can become significantly more expensive for heavy or continuous agentic workloads.

OpenClaw is an open-source project that lets developers build autonomous AI agents capable of running complex, long-running tasks using models like Claude. These agents often consume large volumes of tokens, which has put pressure on Anthropic’s infrastructure during periods of high demand. The company has argued that such third-party tools strain its systems and violate its terms of service, which prohibit using subscription access for external products or services.

Steinberger has suggested the timing feels pointed, especially as Anthropic continues to develop its own Claude Code and agent-related tools. However, Anthropic has maintained that the policy change is about fair usage, system stability, and preventing abuse rather than targeting any specific project.

The brief account suspension of OpenClaw’s creator added fuel to the conversation in developer communities. Many saw it as heavy-handed, while others viewed it as an understandable move by a company trying to manage explosive growth in AI usage without letting costs spiral out of control.

For now, OpenClaw users can still access Claude models, but only by switching to direct API keys and paying per token, rather than relying on their existing Claude subscription. Some developers have already begun exploring workarounds or shifting portions of their workloads to other models.

The episode highlights a broader challenge facing frontier AI companies: how to balance generous consumer subscriptions with the reality that power users and agent frameworks can burn through resources far faster than anticipated. As autonomous agents become more common, expect more friction between model providers and the open-source tools built on top of them.

Anthropic has not issued a detailed public statement on the temporary ban beyond clarifying that it was triggered by automated systems and was not a permanent action against Steinberger or OpenClaw specifically. The quick reversal suggests the company is sensitive to public perception in the tight-knit AI developer community.

Whether this incident leads to lasting changes in how OpenClaw interacts with Claude, or pushes more users toward alternatives, remains to be seen. For the moment, it serves as a reminder that even popular open-source projects operate at the mercy of the large model providers whose APIs they depend on.