OpenClaw: AI Errors and Absurd Refunds

⚡

Key Takeaways

1An OpenClaw user requested a refund for errors on sensitive financial documents.

2Peter Steinberger, creator of OpenClaw, refunded the user $0, the amount initially paid.

3OpenClaw, an open-source AI agent, is used for free, but its errors raise questions about accountability.

💡Why it matters — AI errors in sensitive contexts highlight the challenges of accountability and user expectations.

OpenClaw and the Unexpected Refund Request

The artificial intelligence agent OpenClaw is generating increasing interest among users. Recently, a user attempted to request a refund for errors made by the agent, according to Peter Steinberger, the creator of OpenClaw. Steinberger shared this refund request, which cited errors such as "fabricated data."

The user in question explained that they had used OpenClaw to process sensitive financial documents. However, they had to spend numerous hours verifying and correcting the information provided by the agent. In response, Steinberger refunded the user the exact amount they had paid for the service: $0.

The Risks of AI Autonomy

The question of accountability arises when errors occur with an AI agent. OpenClaw is designed to operate autonomously, performing tasks without human intervention. This autonomy can sometimes lead to mistakes, as illustrated by an incident involving Meta's alignment director, who nearly deleted her entire inbox. For some, these errors are seen as an inherent cost of using AI. For others, they justify a refund request.

Steinberger, who developed OpenClaw before joining OpenAI, posted the refund request on X. The user had configured OpenClaw to process "sensitive financial documents" for over eight hours. Subsequently, they spent hours correcting the generated errors.

The Costly Errors of OpenClaw

Reported errors included incorrect financial figures, invented data, internal contradictions, and erroneous calculations. These mistakes occurred in confidential board documents where accuracy is crucial. The user described these errors as having caused "significant wasted time and frustration," and expressed a desire to be refunded for their session.

As an open-source AI agent, OpenClaw is free to use. Steinberger responded to the refund request by offering the user "a full refund of what they had paid," which amounted to $0. "I felt generous," Steinberger joked, adding that he had "rounded down to zero."

Reactions and Legal Implications

Several commentators speculated that the requester was European, due to the strict regulations in Europe that, according to Steinberger, hinder technological innovation. However, he clarified that the user was actually American.

One commentator pointed out that in many jurisdictions, there are remedies for damages, making the request "less ridiculous than it seems." Steinberger responded by citing a clause from OpenClaw's license agreement.

The license states that "the software is provided 'as is', without any warranty of any kind." The authors cannot be held liable for any claims or damages "related to the software or the use or other dealings in the software."

The Rapid Expansion of OpenClaw

OpenClaw is experiencing rapid growth, with swift adoption by Chinese companies. Lobster hats, the emblem of the community, are gaining popularity, and fans are organizing conferences to share and attend demonstrations.

Meanwhile, Nvidia recently announced its own AI agent, NemoClaw, marking its entry into the bot market.

It is even possible that an AI agent drafted the refund request. "Its claw sent you an email," suggested one commentator. "Very likely," Steinberger replied.