Writing patent application documents or triggering multiple risks using intelligent agents such as OpenClaw.

The State Intellectual Property Office issued a risk warning today, stating that tools such as OpenClaw and other intelligent agents have been exposed to default security vulnerabilities. Using such intelligent agents to draft patent application documents may lead to multiple risks. The State Intellectual Property Office warns of the risk of "technology information leakage": intelligent agents like OpenClaw have high permissions, security vulnerabilities, and plugin poisoning risks, using them to draft application documents may cause leakage of core information such as technical disclosures. Once leaked, the patent application technical solution may lose novelty and be unable to be authorized, or even be preemptively applied for by others, resulting in significant losses to the applicant. The agency will also bear liability for breach of contract and compensation. There is also the risk of "substantial defects": using such intelligent agents to draft application documents may result in "AI illusions", leading to content being logically contradictory and technical features being unclear, making it impossible to obtain protection. The risk of "dishonest application": generating patent applications out of thin air, randomly fabricating, or piecing together content using intelligent agents constitutes dishonest behavior that violates the principles of honesty and trustworthiness in patent applications. For a certain number of such applications, the applicant may face warnings, fines, and other administrative penalties; the agency and agents may face revocation of their practice licenses, cancellation of their qualifications, and in severe cases, be included in the list of serious violations of laws and regulations.