Trademarks, Copyright and the AI Identity Problem in 2026: Protecting a Brand a Machine Can Copy

The trademarks copyright AI identity problem 2026 presents is no longer theoretical, generative models can now clone a brand spokesperson’s voice, replicate packaging design and produce convincing synthetic endorsements in seconds. China’s 2026 Trademark Law amendment, adopted on 26 June 2026 and effective 1 January 2027, compresses opposition windows, strengthens penalties for bad-faith filings and formally recognises online trademark use. Simultaneously, the United States, the European Union and Denmark are each advancing legislative responses to AI-generated identity harms. For foreign brands operating in China, the convergence of these reforms creates an urgent and narrow window to audit registrations, assemble evidence of genuine use and fortify enforcement strategies before the new rules take hold.

TL;DR, Key Actions for Brand Owners

• China deadline. The 2026 Trademark Law amendment takes effect on 1 January 2027. Opposition periods will shorten and bad-faith filing penalties will increase, act now to file defensive marks and compile use evidence.
• Copyright alone is insufficient. Copyright protects original expression and requires human authorship in most jurisdictions; it does not reliably prevent AI-generated impersonation of a brand’s identity, voice or visual style.
• Trademark law fills the gap. Source-identification, consumer-confusion and goodwill protections under trademark law address the commercial harms that AI-generated replicas actually cause.
• Voice and likeness registrations. The US permits registration of distinctive sound marks and is expanding guidance on name, image and likeness (NIL) filings. China’s practice centres on names, logos and figurative marks, but the 2026 amendment’s recognition of online use broadens what counts as commercial activity.
• Multi-jurisdictional enforcement. Combine platform takedowns, China administrative complaints, customs recordation and, where available, criminal referrals into a single coordinated enforcement programme.
• Start this week. Audit your trademark portfolio, collect dated evidence of genuine commercial use in China and engage local counsel to prepare filings before 1 January 2027.

The Problem: When the Infringer Is a Model

Generative AI has fundamentally altered the economics of brand impersonation. What once required a counterfeiter with physical production capabilities now requires only a text prompt and access to a large language or diffusion model. The resulting identity harms are real, measurable and increasingly difficult to contain using traditional IP tools alone.

How Models Reproduce Voice, Face and Style

Modern generative systems can be fine-tuned on small samples of audio, imagery or text to produce outputs that convincingly mimic a specific individual or brand. Voice-cloning models require as little as three seconds of reference audio. Image generators can reproduce trade dress, product packaging and spokesperson likenesses with photographic fidelity. Text models replicate brand tone, slogans and marketing copy. These outputs are not copies of any single copyrighted work, they are synthetic reconstructions of identity, which is precisely why copyright law struggles to address them.

Types of Commercial Harm to Brands

The commercial damage from AI-generated brand impersonation falls into several categories: consumer deception (buyers believing a synthetic endorsement is genuine), brand dilution (unauthorised associations eroding distinctiveness), lost sales (diverted revenue from counterfeit goods marketed through synthetic content) and reputational injury (deepfake endorsements linking a brand to products or causes it has not authorised). Each harm type maps differently to available legal tools.

Copyright, AI Authorship and Its Limits for Identity Protection

Copyright remains an essential IP right, but its architecture, built around human authorship and original expression, creates structural blind spots when applied to AI-generated identity harms. Understanding where authorship AI copyright doctrines break down is critical for brands deciding how to allocate enforcement resources.

Authorship Tests Across Jurisdictions

In the United States, the Copyright Office has maintained that copyright requires human authorship, and works generated autonomously by AI systems are not registrable. China’s courts have taken a more nuanced position, recognising copyright in AI-assisted works where a human author exercises sufficient creative control, but the threshold remains contested. The UK government’s March 2026 report on copyright and artificial intelligence recommended retaining human authorship as the core requirement while acknowledging the need for new mechanisms to address AI-generated content. The European Parliament’s resolution of 10 March 2026 on copyright and generative AI called on the Commission to clarify how existing copyright rules apply to AI training data and outputs, signalling future legislative action without resolving the authorship question.

When Copyright Helps, and When It Does Not

Copyright can be effective when an AI model has been trained on specific copyrighted works and produces outputs that are substantially similar to those works. It may also apply where AI outputs incorporate recognisable elements of a protected work. However, copyright does not protect an individual’s voice, likeness or personal style. It does not address consumer confusion. And it cannot prevent a model from generating original content that merely sounds like, looks like or feels like a brand’s identity without copying any particular protected work. For these harms, trademark law offers a more direct route.

Trademark Law: What It Protects That Copyright Does Not

Trademark law is concerned with source identification and consumer protection, functions that align directly with the harms caused by AI-generated brand impersonation. Where copyright asks whether a work was copied, trademark law asks whether consumers are likely to be confused about the origin of goods or services. This distinction makes trademark enforcement against AI-generated content significantly more practical in cases where the trademarks copyright AI identity problem manifests as marketplace deception rather than artistic reproduction.

Trademark Remedies vs Copyright Remedies

Trademark enforcement offers a broader practical toolkit than copyright in most jurisdictions. In China, trademark owners can pursue administrative complaints through local Market Supervision Administrations, which can act within days to seize goods, impose fines and order cessation. Customs recordation allows interception of infringing goods at the border. Civil proceedings can yield injunctions, damages and destruction of infringing materials. Criminal prosecution is available for serious trademark counterfeiting. Copyright enforcement, while powerful in its own sphere, typically requires slower civil litigation and does not offer the same administrative speed or criminal penalties for identity-based harms.

China Trademark Amendment 2026, What Foreign Brands Must Know

The China trademark amendment 2026, adopted on 26 June 2026 and taking effect on 1 January 2027, represents the most significant overhaul of China’s trademark regime in a decade. For foreign brands concerned about AI-generated impersonation, three changes are particularly consequential: the shortened opposition period, expanded bad-faith filing penalties under Article 54 and the formal recognition of online trademark use as evidence of genuine commercial activity.

Filing Timelines and the Limited Runway for Foreign Brands

Under the current law, third parties have three months from publication to oppose a trademark application. The 2026 amendment shortens this window to two months. For foreign brand owners who rely on monitoring services to identify problematic filings, this compressed timeline demands faster detection and response. Industry observers expect the practical effect to be significant: brands that currently take six to eight weeks to prepare opposition filings will need to streamline their workflows substantially.

The amendment also places greater emphasis on genuine use evidence China trademark holders must maintain. Registrations that cannot be supported by evidence of actual commercial use become more vulnerable to cancellation. For foreign brands, this means assembling a robust evidence portfolio, sales records, marketing materials, e-commerce listings, distributor agreements, well before the 1 January 2027 effective date.

How Article 54 Changes Enforcement Strategy

The revised Article 54 expands the CNIPA’s authority to address bad-faith trademark filings China has long struggled with. Under the amendment, the CNIPA can issue warnings and impose fines directly on applicants found to have filed in bad faith, without requiring the affected brand owner to initiate civil litigation. This administrative route is faster, less expensive and particularly effective against squatters and AI-enabled filing mills that register marks associated with foreign brands. Early indications suggest that the CNIPA intends to use these powers actively, particularly against repeat filers and applications that target well-known foreign brands.

Jurisdictional Map: Registering and Enforcing Voice, Likeness and Style

The legal tools available to protect brand identity from AI replication vary significantly by jurisdiction. Brands operating across borders need a clear map of what can be registered, where, and what enforcement mechanisms are available in each market.

China

China does not currently permit registration of sound marks as a general category, though some distinctive audio logos have been accepted. The primary protection pathways are name marks, figurative marks, trade dress and, under the Civil Code, personality rights (including name and likeness). The 2026 amendment’s recognition of online use strengthens the position of brands that can demonstrate genuine commercial activity through e-commerce platforms, social media and digital advertising. Administrative enforcement through Market Supervision Administrations remains the fastest practical remedy, with investigations typically concluding within weeks.

United States

The USPTO updated its guidance on name, image and likeness (NIL) trademark filings on 29 June 2026, clarifying the circumstances under which personal names and distinctive voices can be registered as trademarks. The register voice trademark US pathway requires proof of distinctiveness and commercial use. Separately, the NO FAKES Act (S.1367), introduced in the 119th Congress, would create a federal right of action against unauthorised digital replicas of an individual’s voice or likeness. If enacted, it would fill a significant gap in federal law, which currently leaves voice and likeness protection largely to a patchwork of state right-of-publicity statutes.

European Union

The European Parliament’s resolution of 10 March 2026 on copyright and generative AI called on the Commission to review how existing copyright rules apply to AI-generated content and training data. The Commission subsequently launched a call for evidence in May 2026 on the review of EU copyright rules. In parallel, Article 50 of the EU AI Act imposes transparency obligations on deployers of AI systems that generate synthetic content, including requirements to label deepfakes. Industry observers expect these parallel tracks to converge into a more comprehensive regulatory framework by 2028.

Denmark

Denmark has advanced draft legislation proposing a copyright-style right over an individual’s face and voice, specifically addressing deepfake law Denmark aims to curb AI-generated impersonation. If enacted, this would represent one of the first European laws to create a standalone personality-based right against synthetic replication, potentially serving as a model for other EU member states.

Trademark Enforcement Against AI-Generated Content: Constructing an Enforcement Route

Effective enforcement against synthetic impersonation requires a coordinated strategy that combines multiple legal and practical tools. No single mechanism, platform takedown, administrative complaint or court action, is sufficient on its own. The most effective programmes layer these approaches and move in parallel across jurisdictions.

Platform Playbook, Evidence Package Checklist

Before initiating any enforcement action, brands should assemble a comprehensive evidence package. This package serves as the foundation for platform takedowns, administrative complaints and court proceedings alike.

• Capture and hash the infringing content. Use forensic tools to download, timestamp and generate cryptographic hashes of the AI-generated content before it is removed or altered.
• Document the source. Record URLs, platform usernames, metadata, model identifiers (where visible) and any promotional context surrounding the infringing content.
• Prepare a comparison exhibit. Place the genuine brand asset alongside the AI-generated replica with annotations identifying the points of confusion.
• Draft a trademark notice. For platform takedowns, include the registration number, registration jurisdiction, a clear statement of the rights infringed and the specific relief requested. For Chinese platforms, submit in Mandarin with supporting certified translations of foreign registrations.
• File an administrative complaint. In China, submit to the local Market Supervision Administration with the evidence package, a power of attorney and certified copies of trademark registrations.

When to Pursue Opposition or Cancellation vs Court Proceedings

Administrative opposition and cancellation proceedings before the CNIPA are typically faster and less costly than civil litigation. They are the preferred route for addressing bad-faith filings and squatted marks. Court proceedings become necessary when the brand owner seeks damages, when the infringement involves complex facts (such as AI model training on brand assets) or when injunctive relief with broader geographic scope is required. In practice, the most effective strategies pursue both tracks simultaneously.

Audit and Remedial Checklist for Brand Owners

With the 1 January 2027 effective date approaching, brand owners should complete the following audit and remedial actions as a matter of priority. Building genuine use evidence China’s amended law will demand is central to this process.

• Portfolio audit. Review all existing trademark registrations in China for gaps, missing subclasses, unregistered stylised marks, figurative elements and trade dress. Identify personal name marks and brand ambassador names that may be vulnerable to AI impersonation.
• Defensive filings. File applications for key defensive marks, including transliterated Chinese-character versions of brand names, stylised logos and any distinctive figurative elements not yet registered.
• Evidence of use assembly. Compile dated evidence of genuine commercial use for every core registration. This includes sales invoices, distributor and licensing agreements, e-commerce listings (with screenshots and dates), advertising materials and social media campaigns.
• Monitoring programme. Implement AI-enabled monitoring across Chinese e-commerce platforms, social media and the CNIPA gazette. Several forensic vendors now offer deepfake detection specifically calibrated for brand impersonation.
• Enforcement templates. Prepare standardised takedown notices, administrative complaint packages and court filing templates in advance, so enforcement can be initiated within days of detection rather than weeks.

Case Studies and Precedent

In a publicly reported 2025 administrative action in Shenzhen, a foreign consumer electronics brand successfully petitioned the local Market Supervision Administration to seize goods marketed through AI-generated product videos that replicated the brand’s trade dress and spokesperson likeness. The administration acted within 14 days of receiving the complaint, confiscating inventory and imposing fines on the distributor. The case demonstrated the speed and effectiveness of China’s administrative enforcement route when supported by a comprehensive evidence package.

In a separate US matter, a multinational cosmetics company used a combination of DMCA takedown notices and federal trademark infringement claims to remove synthetic endorsement videos from major social media platforms. The platforms responded within 48 to 72 hours to the trademark-based notices, significantly faster than to the copyright-based notices, which required more complex originality and ownership arguments. The case underscored the practical advantage of trademark enforcement against AI-generated content in Western markets.

Conclusion, Recommended Roadmap

The trademarks copyright AI identity problem 2026 poses requires immediate, coordinated action. Brands that wait until the China Trademark Law amendment takes effect on 1 January 2027 will find themselves operating under compressed timelines with fewer procedural options. The following six-step roadmap provides a practical framework for the months ahead.

1. Complete a full trademark portfolio audit across all Chinese subclasses by Q3 2026.
2. File defensive applications for unregistered marks, transliterations and figurative elements.
3. Assemble and notarise evidence of genuine commercial use for every core registration.
4. Deploy AI-enabled monitoring across Chinese e-commerce, social media and the CNIPA gazette.
5. Prepare standardised enforcement templates, takedown notices, administrative complaints and court filings, in both English and Mandarin.
6. Engage experienced China trademark counsel through the Global Law Experts network to review your enforcement strategy and expedite filings before the January 2027 deadline.

Sources

1. China National Intellectual Property Administration (CNIPA), Trademark Law (2026 revision)
2. CNIPA, Expert commentary on the 2026 amendment
3. Congress.gov, S.1367 NO FAKES Act (text)
4. USPTO, Name, Image and Likeness (NIL) guidance
5. European Parliament, Resolution on Copyright and Generative AI (10 March 2026)
6. GOV.UK, Report on Copyright and Artificial Intelligence (March 2026)
7. Skadden, “Whose AI Is It Anyway?” (April 2026)
8. Global Law Experts, Denmark Deepfake Law 2026
9. European Commission, Review of EU Copyright Rules (call for evidence, May 2026)
10. UK Parliament / House of Lords, Copyright and AI Report