CAIDP Submits Comments to EU AI Office on Third Draft of General-Purpose AI Code of Practice

The Center for AI and Digital Policy (CAIDP) has submitted  comments to the European AI Office expressing significant concerns with the Third Draft of the General-Purpose AI Code of Practice. As a plenary participant in the drafting process and member of Working Groups 1, 2, and 4, CAIDP warns that the current draft requirements "move further away from the spirit of the EU AI Act, transparency and protection of fundamental rights."

Critical Concerns with the Third Draft

CAIDP's analysis identifies three key areas of concern:

Transparency Undermined

The draft code dramatically limits transparency requirements, with providers only required to disclose "contact details" while any meaningful information regarding data, models, outcomes, and governance is merely optional. This stands in stark contrast to established transparency norms and the AI Act's emphasis on accountability.

Fundamental Rights Downgraded to "Optional"

Most concerning, the latest draft downgrades the assessment of risks to fundamental rights—including discrimination, privacy violations, and threats to democratic processes—to "optional." CAIDP states this approach is "UNACCEPTABLE" and "contradicts both the text and the intent of Article 56(4) of the AI Act."

Copyright Protections Weakened

Rather than reinforcing protections for creators, the Draft Code introduces ambiguity and weakens existing safeguards, creating legal uncertainty and significantly lowering compliance expectations around copyright.

Stronger Implementation Needed

CAIDP emphasizes that "The Code must be the implementation mechanism of the EU AI Act to advance fundamental rights and safety." The Center recommends strengthening the Code by replacing language where "signatories are encouraged" with "signatories commit to" concrete actions.