EU’s AI Act Fails to Safeguard Artists’ Copyright, Creative Groups Protest
AI Act Implementation Sparks Ongoing Protection Hurdles for Creative Sectors
With the European AI Act beginning its enforcement this weekend, industry advocates in the creative field are raising concerns that the new legal framework still falls short of fully shielding their copyrighted works from use in AI training processes.
Key Issues Highlighted by Creators
- Ambiguity surrounding what materials qualify as “training data.”
- Limited enforcement powers to prevent unauthorized use of copyrighted content.
- Insufficient deterrent penalties for breaches of the protection rules.
- Missing clear pathways for licensing copyrighted material for AI training.
Artists Raise Alarm Over Gaps in the European AI Act
When the European Union’s first comprehensive AI regulation began to take effect, arts‑rights organisations argued that the draft still leaves significant shortcomings for creators.
Key Concerns from Creative Defenders
Both the European Composer and Songwriter Alliance (ECSA) and the European Grouping of Societies of Authors and Composers (GESAC) point out that the legislation does not safeguard works used to train generative‑AI systems. They emphasize three main failings:
- Opt‑out ambiguity – There is no clear procedure for artists to refuse the use of their music, literature, or film in AI training.
- Remuneration gaps – The Act lacks mechanisms to ensure creators receive fair payment when their content is exploited by tech firms.
- Transparency deficits – The law does not compel companies to disclose how and where artistic data is employed in AI development.
Industry Implications and Risk Landscape
According to experts, without these protections, artistic assets remain vulnerable as they continue to serve as the backbone of emerging AI models. Artists risk losing control over how their work is represented and monetised.
Statements from Leadership
“The productions of our members should never be utilised without full transparency, explicit consent, and proper remuneration,” said Marc du Moulin, secretary‑general of ECSA. He highlighted that current enforcement of the AI Act fails to meet these basic expectations.
‘Putting the cart before the horse’
The EU’s AI Act: Safeguarding Responsible Artificial Intelligence
Core Purposes
- The legislation aims to guarantee that AI deployments remain safe, transparent, traceable, non‑discriminatory, and environmentally sustainable.
- It introduces a structured risk evaluation system for AI services.
Risk‑Based Categorization
- Four risk tiers are defined: Minimal, Limited, High, and Unacceptable.
- AI solutions flagged as Unacceptable—for instance those that manipulate users or conduct social scoring—are prohibited outright.
- Most generative models, such as large‑language systems, fall under the Minimal risk bracket.
Minimal Risk Compliance
- Even minimal‑risk AI providers must meet certain obligations, including the disclosure of training data summaries that reference copyrighted works.
- Under EU copyright provisions, businesses may use copyrighted content for text and data mining unless a creator specifically reserves their rights.
Addressing the Opt‑Out Gap
- There is no clear mechanism for artists to opt out of having their creations employed in AI training datasets.
- Experts argue that this situation puts the rules behind the contributions they protect.
Voluntary Industry Commitment
- The EU’s General‑Purpose AI (GPAI) Code of Practice invites companies to pledge a comprehensive copyright policy.
- Signatories, including Amazon, Google, Microsoft, and OpenAI, are required to implement safeguards against rights infringement and to establish a dedicated complaint handling portal.
AI providers have to respect copyright laws, the Commission says
AI Act’s Transparency Rules Leave Artists with Limited Recourse
Artists are finding that the new AI Act offers clarity about who has already used their works and when they were accessed. However, the legislation offers little protection for already-processed content, leaving creators unable to claim remuneration for material that has been scraped to train AI systems.
Potential Impact Seen as Non-Retroactive
According to Du Moulin, the AI Act only applies moving forward; it does not empower creators to recover compensation for prior exploitation. “Everything that has already been scraped is effectively a free lunch for AI developers,” he remarked.
Calls for Proper Licensing and Recognition
Adriana Moscono, head of GESAC, reported that attempts by some members to opt out—sending letters and emails to AI firms in order to secure licenses—were met with silence. “There was no answer, and the need to respect copyright was outright denied,” Moscono said.
The European Commission, through spokesperson Thomas Regnier, emphasized that AI operators must uphold copyright holders’ rights during text and data mining. He added that any infringements could be resolved via private settlements.
Key Points from the Latest Statements
- AI Act clarifies usage history but does not cover past scraping.
- Artists seeking licenses have found no response from AI firms.
- Commission stresses respect for copyright during AI training.
- Existing EU copyright laws remain unchanged.
Mandate licence negotiations, groups ask
EU Commission Urged to Clarify AI Copyright Rules
Call for Clearer Guidance from Du Moulin and Moscono
- Du Moulin and Moscono are urging the European Commission to provide an urgent and transparent set of rules regarding opting‑out procedures and the protection of copyrighted content under the forthcoming AI Act.
- Moscono noted that existing guidelines—namely the code of practice, the template, and the guidelines—lack any meaningful bandwidth to improve artists’ standing.
- He warned that without guaranteed enforcement of the Act, artists remain vulnerable.
- Both advocates suggest the Commission could impose an obligation on AI firms to negotiate comprehensive or collective licenses with established artist groups.
Recent Legal Actions Impacting AI Companies
- The German Society for Musical Performing and Mechanical Reproduction Rights (GEMA) has initiated two copyright lawsuits targeting OpenAI (the parent company of ChatGPT) and Suno AI, a music‑generation application.
- Although these suits fall outside the scope of the AI Act, Du Moulin believes their outcomes could influence the extent to which AI firms must adhere to traditional copyright laws.
Commission and EU Court Review of Text‑Mining Exemptions
- The Commission, along with the European Court of Justice—the EU’s highest judicial body—has signaled plans to reexamine the text‑ and data‑mining exemption embedded in the 2019 copyright legislation.
- This review underscores the Commission’s intent to align data‑usage protocols with the evolving AI regulatory framework.
Compliance Timeline for New AI Companies
- All emerging AI enterprises must demonstrate conformity with the AI Act’s provisions by 2026.
- Companies already operational within the EU enjoy a grace period that extends until 2027.
Related: “We Don’t Want Our Music Killing People”: Artists Protest Against Spotify Over AI Warfare Connections

