OpenAI Aligns Safety Practices with EU AI Act's GPAI Code

OpenAI has announced adjustments to its safety and governance practices to align with the European Union's AI Act, specifically the Code of Practice for General-Purpose AI (GPAI). This move positions the company within the evolving regulatory landscape, where transparency, risk management, and accountability are becoming mandatory for high-impact AI systems. The alignment reflects OpenAI's recognition that voluntary commitments must now meet binding standards.

Key Changes in Safety Protocols

The updated framework introduces more rigorous documentation requirements, including detailed descriptions of training data, model capabilities, and limitations. OpenAI will also implement systematic red-teaming and stress-testing procedures that correspond to the GPAI Code's tiered risk categories. These changes aim to demonstrate compliance with the EU's emphasis on systemic risk assessment for models that could influence public safety, fundamental rights, or democratic processes.

By adopting these measures, OpenAI is proactively addressing the EU's demand for explainability and human oversight. The company's revised safety policies now include mechanisms for continuous monitoring and incident reporting, which are central to the GPAI Code. This alignment may set a precedent for other AI developers, as the EU Act's extraterritorial reach affects any provider operating within the European market.

Industry analysts note that OpenAI's move could accelerate the harmonization of global AI safety standards. While the EU Act is not yet fully enforced, early compliance reduces legal uncertainty and builds trust with regulators. However, critics argue that the code's flexibility leaves room for interpretation, and true alignment will require independent audits. OpenAI's decision to publicly detail its alignment steps suggests a strategic effort to shape the emerging governance framework rather than merely react to it.