Why EU AI Code of Practice Has Sparked Tech Industry Debate

The EU’s AI Code of Practice has highlighted distinct perspectives within the technology industry, with some firms advocating for the framework, while others dismiss it as excessive regulation.

This code functions as a practical guide for executing the EU AI Act, renowned as the first global AI legislation aimed at controlling AI systems by assessing potential risks.

The act implements a risk-oriented approach, enforcing tougher conditions on AI viewed as a higher threat to safety, basic rights and societal norms.

Meanwhile, the voluntary Code of Practice mainly tackles general-purpose AI models.

This focus on general-purpose AI models is due to their robustness, encompassing systems from OpenAI, Google and Meta, which can execute various tasks instead of being restricted to specific operations.

These configurations include some of the most sophisticated AI, such as large language models (LLMs) and multimodal systems.

Currently, technology and AI firms are choosing either to sign or to abstain from the code.

Agreement and rejection among industry leaders

The code, a result of collaboration among 13 experts and contributions from more than 1,000 stakeholders like model providers, SMEs, academics and civil society, faces variable reception as the August 2025 enforcement date for mandatory adherence approaches to the EU AI Act.

Microsoft appears poised to endorse the code, as suggested by Vice Chair and President Brad Smith, expressing its anticipation for industry's direct involvement.

“I think it’s likely we will sign. We need to read the documents,” Brad says.

Conversely, OpenAI already endorses the framework, indicating itself as a pioneer among voluntary adopters, along with French developer Mistral.

OpenAI emphasises its obligation to offer capable and secure AI systems, saying: “Signing the code reflects our commitment to providing capable, accessible and secure AI models for Europeans to fully participate in the economic and societal benefits of the Intelligence Age.”

Anthropic, too, has openly stated its intent to comply, aligning with the code’s principles of transparency and accountability.

The company says: "We believe the code advances the principles of transparency, safety and accountability – values that have long been championed by Anthropic for frontier AI development.”

Key framework elements of the EU’s AI code

The EU’s AI Code establishes stipulations across three key areas for developers of general-purpose AI models:

• Transparency
• Copyright compliance
• General-purpose AI with Systemic Risk (GPAISR)

Under these requirements, model providers are required to document their model and dataset details thoroughly.

Compliance with copyright demands detailing of clear policies about data acquisition and utilisation within EU guidelines.

The most sophisticated models, categorised as GPAI with Systemic Risk (GPAISR), face additional safety obligations.

This classification applies to powerful AI systems such as OpenAI’s o3, Anthropic’s Claude 4 Opus and Google’s Gemini 2.5 Pro.

Adhering entities must also release summaries of training data content and carry out comprehensive risk evaluations.

Potential penalties for non-compliance are hefty, with fines reaching up to €35m ($38.5m) or 7% of global annual turnover, whichever is greater.

Specifically for general-purpose AI model providers, the European Commission could levy fines up to €15m ($16.5m) or 3% of worldwide turnover.

Growing industry pushback

Amid approaching deadlines, a coalition of more than 40 major European corporations, including Airbus and ASML Holding, has petitioned the European Commission for a two-year delay in the AI Act’s full implementation.

The Commission has hinted that companies compliant with an accredited Code of Practice will enjoy simpler compliance processes, focusing regulatory scrutiny on code adherence rather than extensive audits of AI systems.

This may incentivise early compliance among firms seeking regulatory certainty with mandatory enforcement beginning in August 2025.

The global alignment of this regulatory framework with initiatives like the G7 Hiroshima AI Process could set a benchmark for international AI regulation.

Despite opposition, Thierry Breton, the former Internal Market Commissioner, insists the framework will advance as planned.