xAI Signs EU AI Safety Rules, Skips Transparency and Copyright Sections

Elon Musk's AI company xAI announced it will sign the safety and security chapter of the European Union's voluntary AI Code of Practice but will skip the transparency and copyright sections.

The EU Code of Practice has three chapters: transparency, copyright, and safety and security. The framework was created by 13 independent experts to help AI companies prepare for the EU's AI Act. Companies can choose to sign voluntarily, which gives them legal certainty under the new rules. Those who don't sign lose these benefits.

The safety and security chapter targets providers of the most advanced AI models. The transparency and copyright chapters apply to all general-purpose AI systems. xAI explained its decision in a post on X.

Other major tech companies have responded differently to the code. Google's president of global affairs Kent Walker said the code encourages “access to secure, high-quality AI tools across Europe” but warned some parts “might hinder innovation.” He cited concerns about copyright law changes, slower approval processes, and risks to trade secrets.

Microsoft President Brad Smith said his company is “likely” to sign the full code. OpenAI and Mistral have already signed all three chapters. Meta Platforms, on the other hand, rejected the entire framework through chief global affairs officer Joel Kaplan, who called it legally uncertain and beyond the AI Act's scope.

The EU's Artificial Intelligence Act entered into force on August 1, 2024. It will create binding rules for high-risk AI systems. General-purpose AI providers must comply by August 2, 2025. Existing models get extra time until August 2, 2027.

The Code of Practice should receive final approval by the end of 2025. It aims to create uniform standards for training data disclosure, security testing, and copyright compliance across the industry.

EU tech chief Henna Virkkunen endorsed the code and urged companies to participate in collaborative compliance. Final approval by member states and the European Commission is planned before year-end to align with the AI Act's rollout schedule.