State Intellectual Property Office Warns: Three Risks in Using AI Agents Like OpenClaw for Patent Writing

China's State Intellectual Property Office recently issued a risk warning, pointing out that AI agent tools like OpenClaw (formerly Clawdbot, Moltbot) have weak default security configurations. Using them to write patent applications may lead to multiple risks.

These risks include technical information leakage (due to excessive permissions or vulnerabilities), substantive defects (AI hallucinations causing logical contradictions or unclear expressions), and unethical applications (fabricating or piecing together content). If these occur, applicants may face patent rejection or fines, and agencies may also bear responsibility.

The office advises applicants to raise risk awareness and choose agency services carefully; agencies should avoid using AI agents for unethical applications to protect clients' rights. If problems are found, complaints can be filed according to law.