The rapid proliferation of artificial intelligence (AI) technologies across virtually every industry has created a complex and evolving landscape of intellectual property (IP) risks. For insurance carriers, understanding the intersection of AI and IP is critical to accurately assessing exposures, developing appropriate coverage products, and managing claims.
The Shifting Landscape of AI-Generated Content
One of the most pressing IP issues involves the legal status of content created by AI systems. Under current U.S. copyright law, copyright protection requires human authorship. The U.S. Copyright Office has consistently refused to register works created autonomously by AI, and recent decisions have reinforced that AI-generated works without sufficient human creative input cannot receive copyright protection.
This creates significant uncertainty for insureds who rely on AI tools to generate marketing materials, software code, product designs, or creative content. If the output lacks copyright protection, competitors may be free to copy it, undermining the commercial value the insured expected to secure. Carriers should therefore consider how existing policies address the loss of expected IP protection and whether disputes arising from the uncertain legal status of AI-generated works fall within or outside available coverage.
Copyright Infringement Risks in AI Training and Output
The training of generative AI models on copyrighted works has become a focal point of litigation. Authors, visual artists, musicians, and news organizations have sued AI developers on the theory that ingestion of copyrighted material into training datasets constitutes infringement. Insureds that develop or deploy AI systems therefore face potential direct, contributory, and vicarious infringement claims. Equally important, AI-generated outputs may substantially replicate protected material from the training corpus, exposing users of those outputs to infringement liability even if they lacked knowledge of the underlying copying.
Recent cases illustrate how unsettled this area remains. In Alcon Entertainment, LLC v. Tesla, Inc., 2026 U.S. Dist. LEXIS 53719, at *6-7 (C.D. Cal. 2026), the court denied Tesla’s motion to dismiss a copyright claim based on allegations that AI image-generation tools were used to create imagery derived from Blade Runner 2049 without authorization. The court reasoned that infringement may plausibly rest not only on similarities in the final output, but also on "literal" or "intermediate" copying that occurs during the AI generation process. The court also declined to resolve fair use at the pleading stage, concluding that factual development was needed, particularly as to commercial purpose and market harm.
By contrast, in Kadrey v. Meta Platforms, Inc., 788 F. Supp. 3d 1026, 1043-1044 (N.D. Cal. 2025), the court granted summary judgment to Meta and held that training large language models on copyrighted books qualified as fair use on the record before it. The court emphasized the transformative character of the use and found no evidence that Meta’s outputs regurgitated the plaintiffs’ works or caused concrete market harm. At the same time, the decision signaled that future plaintiffs could prevail if they developed a stronger evidentiary showing of market dilution or substitution caused by AI-generated works.
Similarly, in In re Mosaic LLM Litigation, 2025 U.S. Dist. LEXIS 153961, at *12-13 (N.D. Cal. 2025), the court addressed discovery issues but underscored the significance of developing evidence concerning licensing negotiations and emerging markets for AI training data. The court declined to hold, as a matter of law, that licensing markets for AI training are irrelevant to fair use, and permitted limited discovery directed to that issue.
For carriers, these cases underscore that AI copyright claims may turn heavily on factual development, making defense costs substantial and early coverage determinations difficult.
Patent Law Uncertainties
The patent landscape presents its own complexities. Courts in the U.S. have held that AI systems cannot be named as inventors on patent applications, meaning that inventions conceived autonomously by AI may be unpatentable. For insureds engaged in AI-driven research and development, this may limit the scope of available patent protection. In addition, AI-related inventions often face scrutiny under subject-matter eligibility doctrines, and many software-implemented claims remain vulnerable to challenges as abstract ideas. Carriers should be aware that insureds may face both difficulty in securing patent rights and difficulty in enforcing them, which can reduce the practical value of an insured’s IP portfolio and create downstream disputes about the scope of coverage for related losses.
Trade Secret Exposure
AI systems often rely on proprietary algorithms, training methodologies, model weights, prompts, and curated datasets that may constitute valuable trade secrets. Yet deployment of AI tools, particularly cloud-based or third-party systems, can create material risks of inadvertent disclosure or misappropriation. Employees or contractors who input confidential material into AI platforms may expose trade secrets to outside vendors or embed them in systems that become accessible to others.
Conversely, AI models trained on improperly acquired confidential data may generate trade secret claims against the developer, deployer, or both. Carriers should therefore assess whether cyber, technology errors and omissions, media, and other policies meaningfully address trade secret exposure in AI settings, including defense obligations, exclusions, and allocation questions.
Trademark and Brand Protection Concerns
AI technologies also intersect with trademark law in meaningful ways. Generative AI tools can produce content that incorporates third-party trademarks, logos, or brand elements without authorization, potentially exposing users to infringement, false endorsement, dilution, or unfair competition claims. AI-powered counterfeiting, impersonation, and deepfake content also threaten brand integrity in ways that traditional trademark enforcement mechanisms may struggle to address. Insureds may therefore face both offensive and defensive trademark issues tied to AI, and carriers should consider how personal and advertising injury, media liability, cyber, and professional liability coverages respond to these emerging scenarios.
Coverage Gaps and Policy Considerations
Commercial General Liability (CGL) and Business Owners Policies (BOP) may provide limited coverage through personal and advertising injury provisions, but that coverage is often sharply constrained in the IP context. A policy may cover injury arising from use of another’s advertising idea in an advertisement that violates a person’s right of privacy or infringes another’s copyright, trade dress, or slogan in an advertisement.
At the same time, many CGL policies contain exclusions for advertising injury that include but are not limited to:
- Knowing violations of rights of another.
- Material published with knowledge of its falsity.
- Prior publication.
- Criminal acts.
- Contractual liability.
- Breach of contract.
- Quality of performance of goods - failure to conform standards.
- Wrong description of prices.
- Infringement of copyright, patent, trademark, or trade secret, subject in some cases to carve-backs for copyright, trade dress, or slogan infringement occurring in the insured’s own advertising.
As a practical matter, these provisions often mean that patent cases are not covered, and that copyright or trademark claims are covered only when the alleged injury is tied closely enough to advertising activity to fall within the policy language. Professional liability, media, cyber, and technology E&O policies may also contain exclusions or limitations that materially restrict coverage in all IP disputes. Accordingly, the allegations in the underlying pleading and the precise causes of action asserted are critical, especially in jurisdictions that apply a strict "eight corners" approach to the duty to defend.
Recent cases show that coverage analysis in IP disputes turns on two recurring issues: whether the alleged injury is tied to advertising activity, and how broadly the governing jurisdiction defines the duty to defend.
Most recently, in Hershey Creamery Co. v. Liberty Mutual Fire Insurance Co., 386 F. Supp. 3d 447, 450 (M.D. Pa. 2019), the court held that the insurer had a duty to defend because the underlying complaint alleged misuse of slogans and advertising ideas in advertising. The court found that those allegations potentially fell within personal and advertising injury coverage, despite the policy’s IP exclusion.
In 2018, courts reached different results depending on how closely the alleged infringement was connected to advertising. In High Point Design, LLC v. LM Insurance Corp., 911 F.3d 89, 94 (2d Cir. 2018), the Second Circuit held that a CGL insurer had a duty to defend trade dress claims to the extent they were based on an advertisement showing a photograph of the insured’s product, even though the policy did not cover claims directed solely at the product or its packaging.
By contrast, in Owners Insurance Co. v. Cruz Accessories, 2018 U.S. Dist. LEXIS 165992, at *12 (D.S.C. Sept. 26, 2018), the court found no duty to defend or indemnify copyright claims because the alleged injury arose from the manufacture and sale of infringing products, not from advertising, and the mere online sale of products did not establish the required causal connection.
Earlier decisions reflect the same divide. In St. Surfing, LLC v. Great American E&S Insurance Co., 776 F.3d 603, 607 (9th Cir. 2014), the Ninth Circuit held that even if trademark allegations could potentially fall within advertising injury coverage, coverage was barred by the prior publication exclusion because the insured’s allegedly infringing branding predated the policy period. In Bridge Metal Industries, LLC v. Travelers Indemnity Co., 559 F. App’x 15, 18 (2d Cir. 2014), however, the Second Circuit held that allegations that advertising and marketing contributed to consumer confusion were enough to trigger a defense under New York’s broad duty-to-defend standards.
The breadth of the governing jurisdiction’s duty-to-defend law can also materially affect the result. In Hudson Insurance Co. v. Colony Insurance Co., 624 F.3d 1264, 1267 (9th Cir. 2010), applying California law, the Ninth Circuit held that once the complaint alleged a potentially covered slogan-infringement claim, the insurer was obligated to defend the entire action. That decision reflects California’s especially broad duty-to-defend principles and shows how the same dispute may yield different coverage outcomes depending on the applicable law.
On the narrower end of the spectrum, in America’s Recommended Mailers, Inc. v. Maryland Casualty Co., 339 F. App’x 467, 469 (5th Cir. 2009), relying on Sport Supply Group, Inc. v. Columbia Casualty Co., 335 F.3d 453, 464-65 (5th Cir. 2003), the Fifth Circuit held that there was no duty to defend where the underlying claims were strictly trademark infringement claims and did not allege use of another’s advertising idea within the meaning of the policy. That reasoning underscores how critical the wording of the underlying complaint can be.
Finally, in Charter Oak Fire Insurance Co. v. Hedeen & Co., 280 F.3d 730, 735-36 (7th Cir. 2002), the Seventh Circuit held that the use of a trademarked name and logo on business letterhead distributed to the public was at least arguably conduct undertaken in the course of advertising, which was enough to trigger a duty to defend.
Taken together, these cases show that coverage in IP disputes often depends on whether the complaint expressly ties the alleged infringement to advertising, whether any exclusion applies, and how broadly the relevant jurisdiction construes the duty to defend.
These issues are likely to become even more complicated as AI-related allegations are layered onto traditional IP theories. A complaint may allege copying during model training, replication in output, misuse of advertising ideas in synthetic content, and dissemination through automated marketing systems, all in the same case. Many legacy insurance products were drafted before the current wave of AI innovation and may not clearly address those scenarios. Carriers should therefore review whether policy definitions, exclusions, endorsements, and defense-cost provisions adequately address AI-specific risks and the likelihood of technically complex, expert-driven litigation.
Underwriting and Risk Management Recommendations
Carriers seeking to underwrite AI-related IP risks responsibly should consider enhanced application questions directed to how the insured uses AI, the provenance of training data, the degree of human review over outputs, and the insured’s governance and documentation practices. Underwriters should also assess whether the insured uses third-party models or vendors, what contractual indemnities are available, and whether internal controls restrict the input of confidential or protected material into external AI systems.
Carriers may also wish to develop AI-specific endorsements or coverage forms that more clearly delineate coverage for claims arising from model training, synthetic outputs, trade secret leakage, deepfakes, and advertising-related misuse of protected content. Encouraging insureds to adopt robust AI governance frameworks, provenance tracking, human-review protocols, and tailored vendor contracts can reduce both the frequency and severity of claims and support more predictable underwriting outcomes.
The intersection of AI and intellectual property presents a rapidly evolving risk environment that demands sustained attention from insurance carriers. Uncertainty in the law, novel theories of liability, and the widespread adoption of AI across industries all contribute to an exposure profile that differs meaningfully from traditional IP risk. Carriers that proactively review policy language, track developing case law, refine underwriting practices, and build expertise in AI-related IP disputes will be better positioned both to serve policyholders and to manage their own risk in this transformative era.
About the authors
Mary-Ellen King is a partner at Lucosky Brookman LLP. meking@lucbro.com
Jean-Marc Zimmerman is a partner at Lucosky Brookman LLP. jmzimmerman@lucbro.com