Anthropic Warns AI Agents Could Create Uncertain Legal Liability

Anthropic says autonomous AI agents could expose it to significant and unpredictable legal claims, Reuters reports, citing the prospectus for the company's stock market debut. Errors, misalignment or security exploits could cause irreversible actions such as deleting data or making financial transactions.

Anthropic said contractual limits on its liability may not be enforceable or sufficient against claims involving autonomous agents. The company also said existing law does not clearly establish whether agents are products, services or another category, or when their actions legally bind the users who deploy them.

US Federal Trade Commission Chairman Andrew Ferguson said developers or users who instruct agents could be liable for harm. He also acknowledged unresolved questions about responsibility when a person uses an AI tool and receives an unexpected result.