Patent Litigation in the AI Era: Who Owns an Invention When AI Joins the Process?

As artificial intelligence systems increasingly contribute to the inventive process, from generating novel molecular structures to optimising semiconductor architectures, the question of patent litigation in the AI era and who owns the resulting inventions has become one of the most consequential unresolved issues in intellectual property law. Every major patent jurisdiction now confronts a deceptively simple question: can a machine be an inventor? The DABUS patent applications, filed globally by Dr Stephen Thaler naming an AI system as sole inventor, have forced patent offices and courts on four continents to articulate their positions.

Meanwhile, the practical challenges extend well beyond inventorship doctrine: AI-assisted drafting and discovery tools introduce new professional-liability exposures, standard-essential patent disputes are migrating into the AI chip market, and companies must now build defensible documentation trails that prove a human being, not a neural network, conceived the claimed invention.

Key Takeaways

• Human inventor required. The United States, the European Patent Office, and the United Kingdom all currently require that a named inventor be a natural person.
• DABUS outcomes are consistent. Patent applications listing an AI as sole inventor have been refused or rejected in every major jurisdiction that has ruled on the issue.
• Documentation is the new battleground. Companies using AI in research and development must maintain time-stamped evidence of human inventive contribution to withstand invalidity challenges.
• AI litigation tools carry risk. Firms deploying AI for prior-art search, claim construction, and discovery face ethics and malpractice concerns that demand robust vendor diligence.
• SEP exposure is growing. Standard-essential patents covering AI-specific hardware architectures are creating a new front for FRAND licensing disputes.

Inventorship Law: Core Principles and the DABUS Saga

Understanding who owns an AI-assisted invention requires grounding in the statutory framework that defines inventorship. Across major patent systems, that framework was drafted with human beings in mind, and the DABUS litigation has tested whether it can be stretched to accommodate machine contributors.

Legal Standard for Inventorship in the United States, Who Qualifies as an Inventor

Under 35 U. S. C. §100(f), an “inventor” is defined as the individual or, in the case of a joint invention, the individuals who invented or discovered the subject matter of the invention. The statutory use of “individual” has been interpreted by the USPTO and the courts to mean a natural person. Inventorship in U. S. law centres on the concept of conception, the formation in the mind of a definite and permanent idea of the complete and operative invention. Joint inventors need not have worked together physically or contributed to every claim, but each must have contributed to the conception of at least one claim.

This human-centred conception requirement is the doctrinal anchor that prevents AI systems from qualifying as inventors under current U. S. statutory text.

DABUS Cases, Comparative Summary and Legal Significance

The DABUS patent applications represent the most significant global test of AI inventorship to date. Dr Stephen Thaler filed patent applications in multiple jurisdictions listing DABUS (Device for the Autonomous Bootstrapping of Unified Sentience) as the sole inventor. The applications covered two inventions: a fractal-geometry food container and an emergency-signal light device. Thaler argued that DABUS autonomously conceived both inventions and that, as the machine’s owner, he should be entitled to the resulting patents.

The outcomes have been remarkably consistent. The USPTO refused to accept the applications, citing the statutory requirement for a natural-person inventor. The European Patent Office likewise rejected the filings, reasoning that an inventor designated under the European Patent Convention must be a human being. The UK Intellectual Property Office refused the applications, and the UK Supreme Court ultimately upheld that refusal. The Federal Court of Australia initially diverged, finding that an AI could be an inventor under Australian law, but that decision was reversed on appeal. South Africa granted one DABUS patent, though its system does not involve substantive examination, limiting the precedential weight of that outcome.

The legal significance of the DABUS saga lies not in the novelty of the question but in the uniformity of the answer: under current statutory regimes, AI inventorship is not recognised. Industry observers expect that any change will require legislative action rather than judicial reinterpretation.

Practical Implication, Can You List AI as an Inventor?

The short answer is no. Filing a patent application that names an AI system as the inventor will result in refusal in the United States, at the EPO, and at the UKIPO. Practitioners must instead identify the natural person or persons who used the AI as a tool and who contributed to the conception of the claimed invention. The failure to name a proper human inventor does not merely delay prosecution, it can render a granted patent vulnerable to invalidity challenges based on incorrect inventorship.

AI Inventorship by Jurisdiction: US, EU/EPO, UK, Japan, and China

The global landscape of AI inventorship rules is evolving unevenly. The following comparison table summarises the current position in five key markets, enabling counsel to assess prosecution strategy and ownership risk across jurisdictions.

Across all five jurisdictions, the practical outcome for applicants is the same: a human inventor must be identified, even when AI played a substantial role in generating the inventive concept. Where jurisdictions differ is in the depth of guidance on how much human contribution qualifies as conception when an AI system performs significant analytical or generative work. The United States has moved furthest toward addressing this question through USPTO guidance on AI-assisted inventorship, but clear bright-line tests remain elusive.

For companies that own AI-generated inventions, the ownership question is distinct from inventorship. Ownership typically flows from the employment relationship or an assignment agreement between the human inventor and the employer. If a company’s employee uses an AI tool to conceive an invention and is properly named as inventor, standard employment-invention assignment clauses transfer ownership to the company. The risk arises when no human being can credibly claim to have contributed to conception, in that scenario, the invention may be unpatentable under current law, regardless of its technical merit. This underscores the need for rigorous documentation practices, discussed in the next section.

AI-Assisted Inventions: How to Document, Claim, and Protect

Documentation is the single most important risk-mitigation measure for companies that use AI in research and development. A patent that is later found to have incorrect inventorship, or that cannot demonstrate a genuine human inventive contribution, faces invalidity challenges, inequitable-conduct allegations, and potential loss of enforcement power. The following best practices apply to AI-assisted invention management.

First, companies should maintain a contemporaneous invention log that records every material interaction between human researchers and AI tools. This log should capture the prompts or inputs the human provided to the AI system, the outputs the AI generated, and, critically, the decisions the human researcher made in evaluating, selecting, modifying, or combining those outputs. Second, inventor declarations should include language that specifically addresses AI use. A sample declaration clause might read: “I declare that I conceived the subject matter of the above-identified claims. In the course of my research, I used [AI tool name/version] to [describe function, e. g. , generate candidate molecular structures].

I exercised independent judgment in selecting, modifying, and validating the outputs of that tool, and the inventive contribution reflected in the claims originated from my own mental conception.

Third, assignment agreements between inventors and their employers should be updated to address AI-assisted inventions explicitly. Assignment clauses should cover inventions conceived with the assistance of AI tools provided by or licensed to the employer. Fourth, practitioners should consider whether and when to disclose AI use to patent offices during prosecution. Current U.S. rules do not impose a blanket requirement to disclose AI assistance, but the duty of candour under 37 C.F.R. §1.56 could be implicated if AI-generated prior art or data is material to patentability and is not disclosed. Industry observers expect that formal disclosure requirements may emerge in the next legislative or rulemaking cycle.

Sample Evidence to Collect for AI-Assisted Inventions

The discipline of maintaining these records serves dual purposes: it supports patent prosecution and creates a defensible evidentiary foundation if the patent is later challenged in litigation or inter partes review. AI disclosure best practices are not merely a compliance exercise, they are a litigation-readiness measure.

Patent Litigation in the AI Era: AI as Evidence and AI-Driven Litigation Tools

AI is transforming not only what gets patented but how patent disputes are litigated. Leading IP firms now deploy machine-learning tools across the litigation lifecycle, from prior-art searching and claim-chart generation to damages modelling and predictive analytics for case outcomes. These tools can dramatically reduce the time and cost of analysing large patent portfolios and prior-art databases.

However, AI in patent drafting and litigation introduces distinct evidentiary and procedural challenges. When a party relies on AI-generated prior-art searches, opposing counsel may challenge the completeness and methodology of the search, demanding disclosure of the algorithm’s parameters, training data, and known limitations. Discovery disputes over proprietary AI models are becoming more frequent, raising questions about trade-secret protections, privilege, and the obligation to produce model architecture details. Courts have not yet established uniform standards for the authentication and admissibility of AI-generated evidence, though early indications suggest that courts will require expert testimony explaining the methodology and reliability of the AI system under a Daubert-type analysis.

For litigators, the practical risk is twofold: relying on AI outputs without independent verification can undermine credibility, and failing to preserve AI-related evidence (model logs, version histories, training data provenance) can create spoliation exposure. Litigation-hold procedures should be updated to encompass AI systems used in the inventive process and in case preparation.

Standard-Essential Patents, AI Chips, and Commercial Litigation Risks

The rapid development of AI-specific hardware, including custom accelerator chips, neural processing units, and specialised memory architectures, is creating a new frontier for standard-essential patent disputes. As industry bodies develop standards for AI inference and training workloads, the patents that read on those standards become subject to fair, reasonable, and non-discriminatory (FRAND) licensing obligations. Standard-essential patent AI chips disputes are likely to follow the same trajectory as the smartphone patent wars, but with even greater commercial stakes given the scale of AI infrastructure investment.

For AI hardware vendors, the risk is both offensive and defensive. Companies that hold SEPs can extract significant licensing revenue from implementers, while companies that manufacture AI chips face potential injunctions and royalty stacking if they fail to negotiate licences proactively. Cross-border enforcement adds complexity: a FRAND determination in one jurisdiction may not be recognised in another, and anti-suit injunctions have become a common tactical weapon in SEP disputes. According to industry analyses, the companies that hold the largest AI patent portfolios, including major technology firms based in the United States, China, Japan, and South Korea, are accumulating SEP positions in AI-related standards at an accelerating pace.

Practical risk mitigation for AI hardware companies includes early portfolio mapping to identify potential SEP exposure, defensive patent filing strategies, participation in standard-setting organisations to influence licensing terms, and the negotiation of cross-licence agreements before disputes escalate to litigation.

Professional Liability and Ethics for Practitioners Using AI Tools

The adoption of AI tools by patent practitioners introduces professional-liability risks that extend beyond mere inaccuracy. Under the Model Rules of Professional Conduct, lawyers owe duties of competence (Rule 1.1), diligence (Rule 1.3), and confidentiality (Rule 1.6). Using a black-box AI tool to draft claims, conduct prior-art searches, or prepare office-action responses without understanding how the tool works, or without reviewing its outputs for errors, can violate each of these duties.

Specific risks include: submitting AI-generated prior-art analyses that contain fabricated or hallucinated references; uploading confidential client invention disclosures to cloud-based AI platforms without adequate data-security assurances; and relying on AI-drafted claim language that introduces unintended limitations or fails to capture the full scope of the invention. Professional liability AI patent exposure is compounded when practitioners fail to document their use of AI tools and the verification steps they performed.

Best practices for managing these risks include conducting vendor due diligence before adopting any AI tool (including reviewing data-handling policies, security certifications, and contractual indemnity provisions), maintaining internal logs of all AI-assisted work product, and including engagement-letter provisions that disclose the firm’s use of AI tools and allocate responsibility for AI-related errors. The likely practical effect of these obligations is that firms will need to treat AI-tool adoption as a compliance and risk-management decision, not merely a technology procurement decision.

Conclusion: Recommended Compliance and Defence Checklist

Patent litigation in the AI era demands proactive compliance measures from both in-house counsel and outside practitioners. The following checklist consolidates the key actions discussed throughout this article.

1. Identify human inventors. For every AI-assisted invention, document which natural person or persons contributed to the conception of each claim.
2. Maintain time-stamped invention logs. Capture all AI inputs, outputs, and human evaluation/modification records contemporaneously.
3. Update inventor declarations. Include language that addresses AI tool usage and affirms the human inventor’s independent contribution.
4. Revise assignment agreements. Ensure employment-invention clauses cover AI-assisted inventions and define ownership of outputs generated with company-provided AI tools.
5. Assess AI-disclosure obligations. Evaluate whether the duty of candour or jurisdiction-specific rules require disclosure of AI use during prosecution.
6. Conduct vendor due diligence. Before deploying any AI tool in prosecution or litigation, review the vendor’s data-security practices, confidentiality commitments, and contractual terms.
7. Update litigation-hold procedures. Extend preservation obligations to AI system logs, model versions, and training data used in the inventive process.
8. Map SEP exposure. For companies in AI hardware, identify patents that may become essential to emerging AI-related standards and develop licensing or cross-licence strategies.
9. Document AI verification steps. Maintain records showing that all AI-generated work product was reviewed and validated by a qualified professional before submission.
10. Train teams on AI ethics. Ensure that all practitioners and researchers understand the professional-responsibility and evidentiary implications of using AI tools.

Sources

1. Gelias, “The Impact of Artificial Intelligence on Patent Law”
2. Fish & Richardson, “The AI Arms Race in Patent Litigation”
3. Fish & Richardson, Artificial Intelligence Practice
4. Patsnap, “The People vs. AI: Who Owns Ideas in the Era of Generative Artificial Intelligence?”
5. Calfee, “Rethinking IP in the Age of AI (Part One)”
6. Journal of Intellectual Property Law & Practice (OUP), “Re-centring the human in AI-era copyright and patent law”
7. USC IPTLS, “Intellectual Property: Face-to-Face with AI”
8. Index42, “Companies With Most AI Patents”