TLDR: Scholarly publishers and other stakeholders have voiced criticism regarding the European Union’s recently adopted General-Purpose AI (GPAI) Code of Practice. While intended to guide compliance with the EU AI Act, concerns persist over its implications, particularly regarding transparency, copyright, and its potential impact on AI development and competitiveness within Europe.
The European Union’s newly published General-Purpose AI (GPAI) Code of Practice, a voluntary framework designed to assist AI model providers in complying with the landmark EU AI Act, has drawn significant criticism from various sectors, including scholarly publishers. The Code, finalized and published on July 10, 2025, after a nearly year-long multi-stakeholder process involving close to 1,000 participants, aims to clarify and standardize compliance approaches for transparency, copyright, and safety/security in AI models.
Despite its stated purpose as a guiding document, the drafting process of the Code was marked by considerable debate and controversy. Critics argue that the Code, while non-legally binding, falls short in several key areas. Long-standing EU discussions surrounding transparency and copyright were reignited during its development. Civil society groups and academics, for instance, advocated for more robust transparency measures, demanding greater insight into the technical workings of AI models and clearer communication to the public. Concerns were also raised that the draft Code might have diluted protections for the fundamental rights of AI system users.
From an industry perspective, a general sentiment among some European companies is that both the EU AI Act and its accompanying Code of Practice are overly stringent. This has led to calls from CEOs of leading European companies for a two-year ‘clock-stop’ on the implementation of the AI Act, citing worries about its complexity and potential adverse effects on Europe’s economic competitiveness. Even major non-EU tech firms have expressed reservations. While Google’s head, Kent Walker, committed to signing the Code, he cautioned that it risks ‘slowing down Europe’s development and deployment of AI.’ Other prominent GPAI model providers like Anthropic and OpenAI have signed the Code, though Meta had not as of recent reports.
Also Read:
- Bloomsbury Considers AI Licensing with Author Opt-In, Sparks Royalty Debate
- Legislators Grapple with Rapid AI Adoption Amidst Evolving Regulatory Landscape
Compliance obligations under the EU AI Act for GPAI models officially commenced on August 2, 2025. The Code of Practice serves as a framework for demonstrating compliance, although adherence does not automatically grant a presumption of conformity. The EU AI Office has indicated a collaborative approach for the initial year: until August 2, 2026, signatories who do not immediately implement all commitments will not be deemed in breach, with the AI Office working with them to achieve full compliance. However, starting August 2, 2026, full enforcement will begin, with potential administrative fines reaching up to €15 million or 3% of a company’s global turnover for non-compliance. A further grace period extends until August 2, 2027, for GPAI models already on the market before August 2, 2025.
