Seven weeks. That’s the runway GPAI model providers have before Article 53 transparency obligations take effect on August 2, and they’re now running it with an active Scientific Panel in place.
The European Commission announced on June 16 that 60 independent AI experts have been appointed to the Scientific Panel, the body established under Article 68 of the EU AI Act. These aren’t advisors in the soft sense. Under Article 68, the Panel has a defined mandate: alerting the AI Office to systemic risks posed by GPAI models, advising on how to classify GPAI systems and evaluate their capabilities, and supporting market surveillance authorities in member states. That’s an operational enforcement role, not a consultative one.
The Commission also constituted the Advisory Forum under Article 67, a complementary body drawing from civil society, academia, and industry. According to the European Commission, the Advisory Forum comprises 174 members. The Commission has not yet publicly announced the Scientific Panel’s Chair and Vice-Chair.
What this changes. Before today, the August 2 deadline carried enforcement authority on paper but lacked the institutional infrastructure to act on it. GPAI providers could comply with the letter of their transparency obligations without worrying much about active scrutiny. That calculus shifts now. The Scientific Panel can formally flag a GPAI model to the AI Office, and once a flag is raised, the AI Office has authority under Article 88 to request access to model documentation, training data summaries, and evaluation results.
The Advisory Forum’s role. The two bodies serve distinct functions. The Scientific Panel is the technical enforcement arm, expert-driven, focused on systemic risk and classification methodology. The Advisory Forum brings in the broader stakeholder perspective: industry concerns, civil society interests, implementation feedback. Think of them as the analytical engine and the political calibration layer. Providers who want to influence how GPAI classification criteria develop should be paying attention to both.
What to Watch
GPAI Provider Compliance, August 2 Readiness
- Technical documentation addressing systemic risk assessment (Article 53)
- Training data summary prepared for potential AI Office review (Article 88)
- Evaluation results documented in format suitable for Scientific Panel review
- Monitor Scientific Panel composition announcement for priority risk categories
What to watch. The Commission hasn’t yet disclosed the full composition of the Scientific Panel, which expertise areas are represented, which institutions the 60 experts come from, whether any members specialize in biosecurity or cybersecurity risk (the domains most likely to trigger systemic risk flags). That roster, when released, will tell GPAI providers a great deal about which risk categories the Panel will scrutinize first. The real question is whether the Panel’s first systemic risk alert lands before or after August 2, an early flag would immediately test how the AI Office exercises its enforcement authority.
TJS synthesis. The appointment of 60 independent experts to the Scientific Panel isn’t a governance formality, it’s the activation of a technical body with formal authority to escalate GPAI concerns to regulators. Providers who’ve been treating Article 53 compliance as a documentation exercise now have a specific audience for that documentation. The Panel will evaluate what you’ve produced. If your technical documentation doesn’t account for systemic risk assessment, the Panel’s first review cycle will expose that gap before you have a chance to close it.