A comparison of the EU's comprehensive AI Act, the UK's 'opt-out' approach, and the US's rights-focused legislative strategy.

The regulation of generative AI in the creative industries is currently fracturing along global lines. While the United States focuses on protecting individual rights through targeted legislation like the NO FAKES and TRAIN Acts, the European Union (EU) and the United Kingdom (UK) have adopted distinct frameworks rooted in comprehensive regulation, risk, and existing copyright directives.

Understanding these different regulatory regimes is crucial, as they determine where AI models are trained and how artists worldwide must enforce their rights.

The US: Rights, Identity, and Litigation

The U.S. approach is characterized by **sector-specific, rights-focused legislation** and extensive reliance on the court system to interpret existing copyright law (Fair Use) and establish new precedents for digital rights.

Key Regulatory Tools:

• **Digital Identity/Likeness:** The **NO FAKES Act** creates a federal Right of Publicity to protect voice and image. This is distinct from EU law, which addresses deepfakes under its broad risk framework.
• **Transparency:** The **TRAIN Act** uses an administrative subpoena to force AI companies to disclose copyrighted training data, directly attacking the "black box" problem.
• **Copyright (Training Data):** The issue is largely settled through litigation. Courts must decide if copying millions of songs for training constitutes **Fair Use** or infringement, establishing a market-driven standard for licensing.

The European Union: Risk, Compliance, and Transparency

The EU's framework is anchored in its comprehensive **AI Act** (the world's first such law) and the existing **Copyright in the Digital Single Market (DSM) Directive.**

The AI Act and Copyright Compliance

The EU AI Act classifies AI systems based on the risk they pose (minimal, limited, high, unacceptable) and applies obligations accordingly. For generative AI (classified as **General Purpose AI - GPAI**), two key obligations directly affect the music industry: [1]

1. **Mandatory Transparency Summary:** GPAI providers must publish a **"sufficiently detailed summary"** of the copyrighted content used for training. This mandatory disclosure requirement is the EU's primary tool for empowering creators to enforce their rights, similar in goal to the TRAIN Act, but broader in scope. [3]
2. **Copyright Opt-Out Compliance:** Providers must implement a policy to respect the reservation of rights expressed by rightsholders under the **DSM Directive**.

The TDM Opt-Out Mechanism

Article 4 of the DSM Directive creates a commercial **Text and Data Mining (TDM) exception** for AI training, but it is conditional: [4]

• AI developers **can** use works for training if they have lawful access (e.g., public internet content).
• **BUT:** Rightsholders can **expressly reserve** or **"opt-out"** their rights using machine-readable means (like specific metadata tags). If a work is opted out, the AI developer must obtain a license to use it. [5]

The United Kingdom: The Push for a Commercial TDM Opt-Out

Post-Brexit, the UK initially explored creating a much broader TDM exception that would have effectively permitted AI training on almost all copyrighted works. However, strong lobbying from the creative sectors led the government to shift its position. [6]

The Proposed Framework

The UK is now proposing to align more closely with the EU by establishing a TDM exception for commercial purposes, but requiring a clear rightsholder **opt-out** mechanism. [6]

• **Goal:** To establish a clear legal basis for AI training, encouraging AI developers to operate in the UK, while simultaneously creating a new, structured licensing market where creators can be paid when they opt-out and the AI developer wants to use the content.
• **Transparency Underpinning:** The proposal explicitly states that the success of the TDM exception depends on **mandatory transparency requirements** for AI developers regarding the works used in their models. [6]

Three Key Differences in Global Policy

The Global Reach of AI Law

A critical point of friction is **extraterritoriality**—the application of a law beyond a country's borders. [7]

• **The EU's "Brussels Effect":** The AI Act attempts to regulate any AI system whose **output enters the EU market**, even if the model was trained outside the EU (e.g., in the US or China). This forces foreign AI developers to comply with EU rules (like the transparency and opt-out requirements) if they want to operate in the massive EU market.
• **US vs. EU:** If US courts rule that training AI is a transformative **Fair Use** (meaning no license is required), while the EU requires compliance with TDM opt-outs, the legal landscape will be deeply fractured, creating complex compliance issues for global AI developers and content distributors.

Conclusion

The global AI divide highlights a fundamental difference in philosophy: the US emphasizes defensive rights and judicial action, while Europe prioritizes systemic regulation and mandatory transparency. For artists operating globally, the future will involve navigating a fragmented legal environment where a work's protection may depend less on where it was created and more on where the AI model was trained and deployed.

References & Further Reading

1. Impact of the EU AI Act on the creative industries - Simkins.
2. EU AI Act: first regulation on artificial intelligence - European Parliament.
3. The EU AI Act and copyrights compliance - IAPP.
4. Generative AI, Copyright and the AI Act - Wolters Kluwer.
5. EU AI Act's Opt-Out Trend May Limit Data Use for Training AI Models - Greenberg Traurig.
6. Copyright and artificial intelligence: Impact on creative industries - House of Lords Library.
7. Artists' Rights in the Age of Generative AI - Georgetown Journal of International Affairs.