IP News
Mitigating Rule 56 Duty of Candor Liabilities in AI-Assisted U.S. Patent Prosecution
•
8 min read
•Views 231
Updated:
//
USPTO administrative guidance issued under 89 FR 25609 confirms that artificial intelligence tools do not alter established practitioner duties of candor and good faith under 37 CFR 1.56. While generative AI models surface diverse inputs, practitioners must apply human verification to categorize and screen information before submitting filings. Indiscriminate reference dumps or unverified submissions in Information Disclosure Statements (IDS) risk violating certification standards under 37 CFR 11.18(b) and duty of candor rules.
Recent legal analysis of the USPTO guidance (89 FR 25609) notes that AI technology does not create a separate duty of disclosure before the USPTO. Instead, obligations under 37 CFR 1.56 depend strictly on whether information shown to prosecution counsel is material to patentability. Generative AI tools combine distinct legal objects inside a single conversational window: authentic prior art, fictitious hallucinated references, contradictory technical facts, and legal conclusions. Under USPTO rules, practitioners cannot satisfy their obligations by submitting unreviewed conversational transcripts or relying on automated IDS population utilities. Submitting unverified, cumulative, or irrelevant references via automated systems risks violating practitioner certification standards under 37 CFR 11.18(b). Additionally, where an AI tool generates alternative claim embodiments, failure to verify that human inventors provided a significant contribution can jeopardize inventorship compliance.
Under 37 CFR 1.56(b), information is material to patentability if it is non-cumulative and either establishes a prima facie case of unpatentability or refutes an applicant's arguments. As shown in Dayco Products, Inc. v. Total Containment, Inc. (329 F.3d 1358), prior-art status under 35 U.S.C. 102 and materiality under Rule 56 are separate legal inquiries; an inconsistent rejection in a co-pending application can be material under Rule 56(b)(2) even if the reference itself is not invalidating. MPEP § 2001.04 and § 2001.06 clarify that material information encompasses diverse facts (including sales, derivation, and litigation statements) regardless of how the information reaches counsel. Furthermore, under 37 CFR 11.104(a)(2), practitioners must reasonably consult with clients regarding the means used to achieve their objectives, requiring appropriate transparency regarding the deployment of AI tools during application preparation.
The USPTO's guidelines create direct operational constraints for prosecution teams and corporate portfolio managers. Practitioners utilizing AI to draft claims or run preliminary searches must manually verify all citations for accuracy, ensuring cited patents exist and removing cumulative material before submitting form PTO/SB/08 under 37 CFR 1.98. For Korean technology companies managing U.S. application portfolios, the operational lesson depends on their procedural posture. When acting as U.S. patent applicants, Korean companies must restrict overseas AI search pipelines from generating bulk reference lists that expose filings to inequitable conduct challenges during litigation. Conversely, when acting as third-party challengers contesting competitors' U.S. patents, Korean entities can inspect prosecution histories for unverified AI outputs or undisclosed AI-generated claim embodiments to challenge patent enforceability. Unlike U.S. practice, where 37 CFR 1.56 imposes an ongoing duty to disclose all known material information under threat of unenforceability, Korean Intellectual Property Office (KIPO) practice does not impose a broad Rule 56-style disclosure duty, relying instead on formal office examination and strict amendment controls under Article 47(2) of the Korean Patent Act.
To mitigate compliance exposure immediately, prosecution counsel should implement a mandatory manual verification protocol for all AI search results prior to IDS submission, logging human verification of every cited document number, publication date, and claim relevance.
The USPTO continues to enforce strict human signature requirements, prohibiting AI systems from executing forms, holding USPTO.gov accounts, or accessing agency databases directly. As patent examiners perform AI-enabled searches using PE2E Search features like "Similarity Search" and "More Like This Document," applicants will face more targeted prior art rejections. IP leaders should establish client-consent protocols under 37 CFR 11.104 and audit outside counsel's software practices. In the interim, applicants should instruct counsel to maintain audited prosecution files that explicitly distinguish human-conceived embodiments from AI-proposed variations to safeguard patent validity against future inventorship challenges.