OpenAI Aligns Safety Practices With EU AI Act as Compliance Phase Begins

OpenAI Signals Compliance With EU AI Act Governance Requirements

According to OpenAI’s blog post published on July 31, millions of Europeans use OpenAI’s tools daily across education, creative work, and business operations. As the EU AI Act transitions into its enforcement phase, OpenAI announced alignment with the regulation’s general-purpose AI framework through endorsement of two codes of practice and publication of updated internal governance structures. The company detailed a Preparedness Framework originally established in 2023 and refreshed in 2025, alongside a newly articulated Frontier Governance Framework designed to operationalize safety and security standards under the EU’s regulatory regime.

Safety Testing and Transparency Mechanisms

OpenAI’s compliance approach emphasizes pre-release safety validation and external oversight. According to the company, it has maintained a Red Teaming Network to involve independent experts in model testing before deployment and publishes system cards alongside major releases. The organization also maintains a public Model Spec document outlining how it shapes model behavior—a transparency mechanism aligned with the EU AI Act’s disclosure requirements. OpenAI reports that its Preparedness Framework structures risk identification, evaluation, and management for advanced systems, while the Frontier Governance Framework maps these internal practices directly onto emerging legal obligations, including requirements for risk assessment, safeguards, model reporting, security incident response, and ongoing expert input.

Institutional Collaboration Beyond Company-Level Action

OpenAI framed responsible AI governance as requiring coordination beyond individual organizations. The company participates in the Frontier Model Forum and maintains collaborative relationships with US CAISI (Center for AI Security and Innovation) and UK AISI (AI Safety Institute), according to the blog post. These partnerships support third-party evaluation practices and shared safety research—positioning industry-wide standardization as prerequisite to effective regulation.

Why This Matters

For EU regulators and member states: OpenAI’s published frameworks provide a template for assessing general-purpose AI model compliance. The timing—during the AI Act’s transition to active enforcement—signals vendor readiness and may influence how regulators calibrate audit and reporting requirements.

For European enterprises and governments: Transparent governance documentation reduces procurement uncertainty. Organizations evaluating OpenAI tools can reference the Preparedness and Frontier Governance frameworks to assess alignment with internal risk policies and regulatory obligations.

For the broader AI industry: OpenAI’s dual-track approach—internal governance plus industry collaboration—suggests that future AI regulation will reward documented processes and third-party validation over assertion alone. Competing vendors may face pressure to match transparency and external oversight standards.