Anthropic to sign the EU General-Purpose AI Code of Practice

Anthropic intends to sign the European Union's General-Purpose AI Code of Practice to promote transparency, safety, and accountability in frontier AI development. This move aligns the company's internal safety standards with EU regulatory frameworks to enable the deployment of AI while mitigating systemic risks.

Economic Impact and Innovation in Europe

Flexible safety standards are viewed as a mechanism to preserve innovation while enabling the broader deployment of AI across the European Union. According to a cited analysis, AI has the potential to contribute more than a trillion euros per year to the EU economy by the mid-2030s.

Anthropic identifies several sectors where AI is already delivering tangible results in Europe:

• Pharmaceuticals: Novo Nordisk is using AI to accelerate breakthrough drug discovery.
• Legal Services: Legora is transforming legal workflows.
• Public Governance: The European Parliament is using AI to expand citizen access to decades of archives.

Transparency and Risk Mitigation Frameworks

The EU Code of Practice establishes a baseline for AI safety through mandatory Safety and Security Frameworks. These frameworks require companies to document how they identify, assess, and mitigate systemic risks, including catastrophic risks related to Chemical, Biological, Radiological, and Nuclear (CBRN) weapons.

Anthropic notes that these EU requirements build upon its own existing Responsible Scaling Policy (RSP), which provides a structured approach to managing risks as model capabilities increase.

The Role of Flexibility and Collaborative Governance

Because AI technology evolves rapidly, Anthropic emphasizes that regulatory policies must remain flexible and adaptive. The company highlights its own iterative approach to safety, noting that its Responsible Scaling Policy has been refined multiple times based on implementation insights, such as clarifying the scope of actors for the ASL-3 Security Standard based on updated threat models.

To maintain this flexibility, Anthropic advocates for a collaborative ecosystem involving:

• The EU AI Office: To ensure the Code remains responsive to emerging technologies.
• Third-party organizations: Groups such as the Frontier Model Forum are cited as critical for establishing common safety practices and evaluation standards that translate technical insights into actionable policy.