This Article Was Written Entirely by AI — Welcome to Synthetic Patent Litigation

1. AI Is Already Reshaping Patent Litigation Workflows: Generative AI is increasingly being used to assist with legal research, prior-art searching, document review, claim charts, deposition preparation, and other patent litigation tasks. 
   
   
2. Disclosure Obligations Remain Unsettled, but Professional Duties Do Not: Courts, the USPTO, and bar associations continue to take varied approaches to AI disclosure, while existing duties of competence, confidentiality, supervision, and verification remain firmly in place. 
   
   
3. Patent Litigators Should Be Prepared to Explain AI Use: Lawyers should become “disclosure-ready” by understanding what AI tools were used, what information they received, and how their output was reviewed and verified.

This article was written entirely by ChatGPT 5.5 without a human directly editing the model’s output. The authors provided the prompt (below) and reviewed the citations, but did not rewrite the prose, reorder the sections, or edit anything by hand. That is the point of this article. It is intended as a demonstration for you: polished legal prose can now be generated quickly enough, and plausibly enough, that readers may not know when AI has played a meaningful role unless that role is disclosed.

Patent litigation is an especially useful setting for that demonstration. It is technical, document-heavy, deadline-driven, and expensive. It also contains many tasks that AI tools can approximate quickly: summarizing file histories, comparing claim language to product documents, generating deposition outlines, distilling expert reports, searching for prior-art vocabulary, and turning large records into usable prose.[1]

Responsible patent litigators should be thinking about those tools now. Chief Justice Roberts has observed that legal research may soon be “unimaginable” without AI, while also warning that AI can invade privacy interests, dehumanize the law, and cannot replace legal determinations that require human judgment.[2] The challenge for patent litigators is to take both sides seriously: AI may improve litigation, but only if lawyers can govern it, verify it, and explain it when explanation becomes necessary.

AI Is Becoming Ordinary—and Harder to See

Generative AI is no longer exotic. ABA Formal Opinion 512 recognizes that these tools may assist lawyers with legal research, due diligence, document review, regulatory compliance, and drafting legal documents. Survey data point the same way: the ABA reported AI use in 30.2% of respondents’ offices, rising to 47.8% at firms of 500 or more lawyers; Thomson Reuters reported that about half of surveyed professionals had used generative AI in some fashion and that 95% believed it would become central to organizational workflows within five years.[3]

The tools are also becoming less visible. They are being embedded in legal-research platforms, e-discovery systems, document-management tools, contract platforms, and law-firm workflows.[4] A lawyer may not open a stand-alone chatbot and ask it to “write a brief,” but may still use AI inside tools already embedded in the workflow. For patent litigators, the better questions may not be whether “AI” was used, but what tool was used, what data it received, what task it performed, and how the result was checked.

Disclosure Rules Are Patchy, but Duties Are Familiar

There is no universal rule requiring lawyers to disclose every use of AI. ABA Formal Opinion 512 instead ties AI use to familiar duties of competence, confidentiality, communication, supervision, candor, meritorious claims, and reasonable fees, while recognizing that the technology is a “rapidly moving target.” [5] A grammar edit is not the same as AI-generated legal research; a private, enterprise tool used on public material is not the same as a public chatbot receiving confidential source code or license data. Client disclosure may be required where the client asks, outside counsel guidelines require it, confidential information will be entered, fees are affected, or AI output will influence a significant decision.[6]

The USPTO has taken a measured approach too: AI tools are not prohibited, but papers must still be signed by a natural person, signers must conduct a reasonable inquiry, and duties of candor and disclosure remain in force.[7] Courts are building a patchwork. Some local rules and standing orders require certifications or disclosure; others emphasize that existing attorney-signature and Rule 11 obligations already apply.[8] That unevenness makes superficial disclosure less useful. “AI was used” tells a court or client little unless the lawyer can say what the tool did, what it saw, and how the output was verified.

Hallucinations Are Only the Easiest Problem

Early sanctions cases made the AI problem easy to caricature. In Mata v. Avianca, lawyers submitted nonexistent AI-generated authorities and were sanctioned after the court found that a “fake opinion” had been submitted as a real authority.[9] The Second Circuit later put the point bluntly in Park v. Kim: “an attempt to persuade a court or oppose an adversary by relying on fake opinions is an abuse of the adversary system.” [10] Since then, courts have sanctioned lawyers for hallucinated authorities, defective attempted cures, and failures to cite-check AI output.[11]

Recent appellate and patent-specific examples make the issue harder to dismiss as a novelty. In Fletcher v. Experian Information Solutions, the Fifth Circuit sanctioned counsel after finding that counsel used AI to draft a substantial part of a brief and failed to verify false quotations, citations, and legal assertions.[12] In Lexos Media IP, LLC v. Overstock.com, Inc., a patent case, the court sanctioned not only the lawyer who used the tool, but all five attorneys who signed filings containing defective AI-generated material.[13] For patent lawyers, the lesson is that signature, supervision, and verification obligations remain personal even when the flawed text came from a machine.

But the harder problem is the real case cited for the wrong proposition, the real patent reference summarized in a way that changes what it teaches, or the real claim term paraphrased in a way that moves the argument. Researchers found that leading AI legal research tools still hallucinated between 17% and 33% of the time.[14] Patent litigation heightens that risk because the consequential errors often are subtle: a prosecution-history statement stripped of context, an expert opinion summarized without its assumptions, a prior-art reference described one limitation too broadly, or a damages theory made to sound more apportioned than it is. Confidentiality and data control are separate concerns: public, enterprise, vendor, and firm-hosted tools may differ in retention, training, access, auditability, and contractual protections.[15]

Becoming Disclosure-Ready

A practical takeaway for patent litigators may be to become disclosure-ready. That does not mean every AI-assisted task should be announced to an adversary, or that courts should require input logs for grammar edits. It means lawyers may want tools, policies, and matter records that can answer basic questions if a court, client, agency, or co-counsel asks: the tool used, the task performed, whether confidential or protected material was entered, and how the output was verified.[16]

Verification will differ by task. Legal citations should be checked against primary sources. File-history summaries should be checked against Office actions, amendments, interviews, and applicant statements. Claim charts should be checked, limitation by limitation, against source material. Damages drafts may require review of licenses, product mapping, royalty bases, apportionment, and expert assumptions. The point is preserving the lawyer’s ability to explain what work was done by a machine and what judgment was exercised by counsel. Clients may want efficient, governed, explainable use, and firms may therefore want policies specific enough to be useful.[17]

Conclusion

Synthetic litigation is already here, but the rules are not settled: courts, bars, and the USPTO are moving unevenly, while the technology moves quickly enough that a policy written for yesterday’s chatbot may not answer tomorrow’s agentic workflow.[18] That uncertainty is not a reason to wait. It is a reason to build habits before the hard question arrives in a live dispute.

Lawyers may not need to disclose every AI-assisted task, but they may increasingly need to know what happened well enough to explain it. AI can make patent litigation cheaper, faster, and better, but only if lawyers keep control over the inputs, verification, confidentiality consequences, and final legal judgment. The better professional posture is readiness: to use the tools where they help, avoid them where they do not, and account for their use when reasonably asked. That discipline may matter later.

This article was written by AI. The next one you read may very well be, too. But it might not tell you.

Prompt:

Please draft a polished article for IP Litigator titled “This Article Was Written Entirely by AI — Welcome to Synthetic Patent Litigation,” authored by Alex Harding and Dan Cooley of Finnegan. Use a serious, practical, slightly provocative tone appropriate for experienced patent litigators, in-house counsel, and judges. The article should be forward-looking but careful, emphasizing professional responsibility rather than hype. Use short sections and endnotes rather than inline citations. Include this prompt verbatim at the end and keep the article to 3 pages before endnotes. Carefully cite-check each citation you make and use twice as many tokens (out of your total budget) as you would normally allocate to doing so.

The article should disclose at the outset that it was generated using an advanced generative AI tool, with the authors providing the prompt, reviewing the output, checking citations, and exercising legal judgment before publication. Make clear that the point is not to celebrate AI-generated legal writing or to suggest that lawyers can outsource judgment. Instead, the article should use its own creation as a practical demonstration of how quickly AI-assisted legal work is entering mainstream litigation practice, and how difficult it may become for readers, courts, and opponents to know when AI has played a role.

Focus on patent litigation specifically. Explain that AI tools may soon assist, or already are assisting, with legal research, brief drafting, prior-art searching, file-history review, claim-chart generation, discovery review, privilege logs, deposition preparation, expert reports, demonstratives, and Federal Circuit issue spotting. At the same time, emphasize that these tools create risks involving hallucinated authority, incomplete reasoning, confidentiality, privilege, work-product protection, supervision, bias, and overreliance.

Please also spend some time addressing our current uncertainty over disclosure obligations. Discuss ABA Formal Opinion 512, USPTO guidance on AI use, court orders or local rules requiring AI disclosures or certifications, and sanctions cases involving hallucinated or defective AI-generated filings. The discussion should be balanced and should not take too strong a stance too early; disclosure should not be treated as a universal requirement in every circumstance, but lawyers should recognize that courts, clients, agencies, and co-counsel may increasingly ask whether AI was used and how.

Develop a practical takeaway that patent litigators may want to become disclosure-ready. This should be suggestive rather than prescriptive. Possible readiness steps may include identifying the tool used, the task performed, whether confidential or client material was entered, whether the tool was public or enterprise-protected, what human review occurred, how citations were verified, and whether the final work product reflects counsel’s independent legal judgment.

End with the sentence: “This article was written by AI. The next one you read may very well be, too. But it might not tell you.” Technically, giving a direct quote may be fairly our own words, but hey. If our reader has read down this far in the prompt, they’re clearly deserving of hearing at least a few real ones from us.

Endnotes

[1] See ABA Comm. on Ethics & Prof’l Responsibility, Formal Op. 512, at 1-5 (2024) (describing generative-AI use in legal research, due diligence, document review, regulatory compliance, and drafting letters, contracts, briefs, and other legal documents); U.S. Patent & Trademark Office, Guidance on Use of Artificial Intelligence-Based Tools in Practice Before the USPTO slides 7, 10-14 (May 15, 2024) (describing AI use in prior-art searching, content generation, and analysis of examiner behavior).

[2] Chief Justice John G. Roberts, Jr., 2023 Year-End Report on the Federal Judiciary 5-6 (Dec. 31, 2023).

[3] See ABA Formal Op. 512, supra note 2, at 1-3; Mark Calaguas, 2024 Artificial Intelligence TechReport, ABA Legal Tech. Resource Ctr. (Apr. 25, 2025); Thomson Reuters Institute, 2025 Generative AI in Professional Services Report 1-3, 30 (2025).

[4] See Mike Scarcella, Anthropic Expands Claude’s AI Tools for Law Firms, Lawyers, Reuters (May 12, 2026); Reuters, Legal Software Firm Harvey Valued at $11 Billion in Latest Funding Round (Mar. 25, 2026) (reporting by Pragyan Kalita and Prakhar Srivastava).

[5] ABA Formal Op. 512, supra note 2, at 1, 3-5, 8-12.

[6] Id. at 6-9, 11-12; see Fla. Bar Ethics Op. 24-1 (Jan. 19, 2024); D.C. Bar Ethics Op. 388 (Apr. 2024); State Bar of Cal. Standing Comm. on Prof’l Responsibility & Conduct, Practical Guidance for the Use of Generative Artificial Intelligence in the Practice of Law 1-4 (2026).

[7] Guidance on Use of Artificial Intelligence-Based Tools in Practice Before the United States Patent and Trademark Office, 89 Fed. Reg. 25,609, 25,612-15 (Apr. 11, 2024).

[8] E.D. Tex. Gen. Order 25-07, at 2, 4-5 (Oct. 31, 2025) (amending Local Rules CV-11(g) and AT-3(m)); N.D. Tex. L. Civ. R. 7.2(f)(1), (3) (eff. Sept. 2, 2025); see Standing Order Re: Artificial Intelligence in Cases Assigned to Judge Michael M. Baylson (E.D. Pa. June 6, 2023); Hon. Aaron D. Maslow, Part Rules § C15 (N.Y. Sup. Ct., Kings Cty.); Fletcher v. Experian Info. Sols., Inc., No. 25-20086, slip op. at 3-4 (5th Cir. Feb. 18, 2026).

[9] Mata v. Avianca, Inc., 678 F. Supp. 3d 443, 449-51, 461, 466 (S.D.N.Y. 2023).

[10] Park v. Kim, 91 F.4th 610, 615 (2d Cir. 2024).

[11]Wadsworth v. Walmart Inc., 348 F.R.D. 489, 493, 495-99 (D. Wyo. 2025); Lacey v. State Farm Gen. Ins. Co., No. CV 24-5205 FMO (MAAx), 2025 WL 1363069, at *1, *5 (C.D. Cal. May 5, 2025).

[12] Fletcher, No. 25-20086, slip op. at 2-5.

[13] Lexos Media IP, LLC v. Overstock.com, Inc., No. 22-2324-JAR, ECF No. 218, at 1, 4, 19-20, 25-26, 35 (D. Kan. Feb. 2, 2026); see also Sara Merken, Judge Fines Lawyers $12,000 over AI-Generated Submissions in Patent Case, Reuters (Feb. 3, 2026).

[14] Varun Magesh et al., Hallucination-Free? Assessing the Reliability of Leading AI Legal Research Tools, 22 J. Empirical Legal Stud. 216, 216-42 (2025), https://doi.org/10.1111/jels.12413 ; ABA Formal Op. 512, supra note 2, at 3-4.

[15] ABA Formal Op. 512, supra note 2, at 6-7; 89 Fed. Reg. at 25,615-16.

[16]See Fed. R. Civ. P. 11(b), 26(g)(1); 37 C.F.R. § 11.18(b) ; 89 Fed. Reg. at 25,612-15; ABA Formal Op. 512, supra note 2, at 3-5.

[17] Thomson Reuters Institute, 2025 Generative AI in Professional Services Report 22-24 (2025); ABA Formal Op. 512, supra note 2, at 9-10; Lexos, No. 22-2324-JAR, ECF No. 218, at 31-35; see Nat’l Inst. of Standards & Tech., Artificial Intelligence Risk Management Framework: Generative Artificial Intelligence Profile 1-3 (NIST AI 600-1, July 2024).

[18] See Fletcher, No. 25-20086, slip op. at 3-4; 89 Fed. Reg. at 25,612-15; ABA Formal Op. 512, supra note 2, at 2-5; Thomson Reuters Institute, 2025 Generative AI in Professional Services Report 27-28 (2025).

https://prod.resource.cch.com/resource/scion/document/default/(%40%40PZG01%20WK-JSTORY20260701-9)4066e96b7a2840c695c27b196264becf?cfu=Legal&cpid=WKUS-Legal-Cheetah&uAppCtx=cheetah