OpenClaw developer Peter Steinberger shared on social media early Friday that maintaining compatibility with Anthropic's models will likely become more challenging in the future. He posted a screenshot showing his account had been suspended by Anthropic due to "suspicious" activity. The suspension was short-lived; within hours, after his post gained significant attention, Steinberger confirmed his access had been restored.

The comment section of his post, which included hundreds of responses, featured speculation and theories, particularly given Steinberger's current employment at OpenAI, a competitor to Anthropic. An Anthropic engineer also commented, clarifying that the company has never banned anyone for using OpenClaw and offered assistance. It remains unclear if this intervention directly led to the account reinstatement.

This incident followed an announcement from Anthropic last week stating that subscriptions to its Claude model would no longer support "third-party harnesses including OpenClaw." Users of such tools must now pay separately based on usage through Claude's API. This move has been informally described as imposing a "claw tax." Steinberger stated he was adhering to this new policy and using the API, yet his account was still suspended.

Anthropic explained the pricing adjustment was necessary because subscriptions were not designed to accommodate the "usage patterns" associated with tools like OpenClaw. These tools can demand more computational resources due to features like continuous reasoning loops, automated task retries, and integration with various third-party services.

Steinberger publicly questioned the timing of these changes. He suggested a pattern where Anthropic first incorporates popular features into its own proprietary agent, Claude Cowork, and then restricts open-source alternatives. He may have been referencing features like Claude Dispatch, which allows remote agent control and task assignment, released shortly before the OpenClaw pricing update.

His frustration was evident again on Friday in response to a comment implying his job choice at OpenAI over Anthropic was a factor. Steinberger replied sharply, "One welcomed me, one sent legal threats."

When asked why he uses Claude instead of his employer's models, he clarified his usage is strictly for testing purposes to ensure OpenClaw updates remain functional for Claude users. He distinguished between his role at the OpenClaw Foundation, aimed at supporting all model providers, and his position at OpenAI, focused on future product strategy.

Others noted that testing Claude is essential because it remains a preferred model for many OpenClaw users compared to ChatGPT. Responding to a comment about the recent pricing change, Steinberger simply stated, "Working on that," offering a hint about his strategic work at OpenAI.

Steinberger did not respond to a request for further comment.