Overview

• The European Commission published detailed rules for AI models deemed to pose systemic risks, including mandatory risk assessments, adversarial testing, incident reporting and cybersecurity measures.
• A voluntary code of practice for general-purpose AI models outlines transparency steps such as technical documentation, copyright compliance and content-use summaries.
• Companies that sign the code can secure legal certainty and reduced administrative burden, and those that bypass it face closer scrutiny by the EU’s AI Office.
• OpenAI and French start-up Mistral have signed the code, Microsoft’s Brad Smith said the firm is likely to join, and Meta publicly declined to participate.
• Providers of systemic-risk and foundation models must comply by August 2 or risk fines from €7.5 million or 1.5% of turnover up to €35 million or 7% of global revenue.