Patent Drafting by AI Agents Raises Official Concerns
China's National Intellectual Property Administration (CNIPA) has issued a formal risk advisory, highlighting potential dangers associated with using artificial intelligence agents to prepare patent application documents. The notice points out that certain popular AI tools have been found to have weak default security configurations, which could expose sensitive technical data and application details to unauthorized access.
A Spectrum of Identified Risks
Beyond immediate security concerns, relying on automated tools for patent drafting introduces several other layers of risk:
- Legal Compliance Issues: Machine-generated text may fail to meet the stringent requirements of patent law regarding novelty, inventiveness, and utility, potentially leading to rejection.
- Inadequate or Inaccurate Disclosure: AI may not fully grasp or correctly articulate complex technical nuances and the essence of an invention, resulting in poorly defined protection scope.
- Ownership and Confidentiality Challenges: Submitting undisclosed invention concepts through third-party AI platforms carries the risk of data retention or misuse.
CNIPA underscores that patent applications are legal documents foundational to securing intellectual property rights. Their drafting quality is paramount. While technology can enhance efficiency, the core tasks of describing inventions and crafting claims still require expert human judgment and expertise.
Moving Forward: Recommendations and Outlook
Industry experts recommend treating AI tools as assistants for research or formatting, not as replacements for qualified patent professionals. For documents containing core technical secrets, ensuring process security and control is paramount. The future may necessitate developing specific review standards and safety protocols for AI-generated content within the intellectual property domain.