OpenAI aligns safety practices with EU AI Act GPAI Code

OpenAI has said its safety, security, and transparency practices are aligned with the EU AI Act’s GeneralPurpose AI Code of Practice as enforcement nears. The company says it has endorsed the GPAI Code and the EU’s Code of Practice on Transparency of AIGenerated Content. It points to existing measures such as prerelease testing, system cards for major launches, redteaming, and a public Model Spec as evidence of its current approach. OpenAI also described two internal frameworks that guide how it handles risk and compliance: its Preparedness Framework and a Frontier Governance Framework. The company says these cover risk assessment, safeguards, reporting, security, incident response, and involvement from external experts. It also highlighted work on provenance tools like Content Credentials and SynthID, plus a cyber programme aimed at giving vetted defenders access to advanced models in the EU.