On November 16, 2026, the European Commission’s two senior advisory bodies—the Group of Chief Scientific Advisors and the European Group on Ethics in Science and New Technologies—issued an unprecedented joint statement warning that traditional public policy guardrails cannot govern rapidly advancing artificial intelligence. Warning that increasing autonomy demands urgent public oversight, the advisors called on the European Union to establish sovereign capacity to evaluate complex AI models rather than delegating safety assessments to commercial technology firms, while demanding rigorous enforcement of the EU AI Act.
Four core recommendations for governing frontier artificial intelligence
The joint statement, titled ‘The next phase of artificial intelligence – Governing for safety, security and human agency,’ sets out four core principles to guide European policymakers through the expansion of generative and agentic systems. Rather than presenting abstract ethical aspirations, the advisory bodies focused on structural governance mechanisms to prevent technological risks from outpacing administrative capacity.
The four primary recommendations delivered to the College of Commissioners comprise:
- Keep existing rules strong: Safety, cybersecurity, and human agency must remain mandatory prerequisites for AI development. The advisors explicitly warn against regulatory backsliding, noting that political hesitation risks weakening statutory obligations under Regulation (EU) 2024/1689, while lobbying efforts to dilute compliance deadlines must be resisted.
- Calibrate development and deployment to degrees of danger: The pace at which foundation models and autonomous systems are released into circulation must correlate with potential hazards. Where models pose severe cybersecurity, biosecurity, or critical-infrastructure risks, deployment must be contingent on verified mitigation rather than post-incident response.
- Build independent public evaluation capacity: Emphasizing technological sovereignty, the advisors urge the EU to develop public testing facilities. The bloc cannot remain dependent on self-attestations or proprietary safety benchmarks supplied by commercial developers when assessing frontier model capabilities.
- Establish an ecosystem for continuous learning: Regulatory frameworks and scientific risk assessments must evolve dynamically alongside advancing model capabilities. Static compliance checklists are obsolete; authorities must create adaptive monitoring loops that ingest empirical evidence from operational deployments.
Summary of advisory recommendations and institutional implications
To assist research administrators and policy analysts in evaluating the joint statement, the following table summarizes the core advisory recommendations, their technical mechanisms, and their practical implications for the research community.
| Advisory Pillar | Governance and Technical Mechanism | Direct Impact on Universities and Research Offices |
|---|---|---|
| 1. Statutory Rule Enforcement | Strict implementation of Regulation (EU) 2024/1689; resistance against legislative dilution or compliance pauses. | Institutional procurement must verify supplier compliance with Article 53 general-purpose AI obligations and copyright transparency mandates. |
| 2. Calibrated Deployment | Pacing commercial release to demonstrated risk mitigation and systemic threshold evaluations. | Research labs training frontier models must implement pre-deployment safety cases before releasing autonomous weights. |
| 3. Independent Model Evaluation | Establishing public European benchmarking infrastructure and model audit testbeds outside commercial control. | Universities gain access to independent validation facilities, reducing reliance on proprietary vendor safety disclosures. |
| 4. Continuous Learning Ecosystem | Adaptive risk governance, ongoing empirical vulnerability tracking, and dynamic safety frameworks. | Institutional Review Boards (IRBs) and ethics committees must establish continuous oversight for deployed agentic tools. |
Background: An unprecedented joint mandate from science and ethics bodies
The release of the joint statement marks a rare procedural convergence between two distinct advisory institutions within the European Commission. The Group of Chief Scientific Advisors (GCSA), established under the Scientific Advice Mechanism (SAM), comprises seven eminent researchers appointed in their personal capacity to deliver evidence-based policy counsel. The European Group on Ethics in Science and New Technologies (EGE) operates as an independent advisory body examining where scientific discovery intersects with fundamental European rights.
While both bodies have previously advised the Commission on computational technologies—delivering opinions on artificial intelligence in crisis management and publishing detailed guidance on the successful and timely uptake of artificial intelligence in science in the EU—authoring a collective joint statement represents an extraordinary intervention. Officials within the Directorate-General for Research and Innovation confirmed that the move reflects mounting concern that commercial market pressures threaten to bypass fundamental safety reviews.
Regulatory landscape: Transitioning from rule-making to independent enforcement
The joint statement arrives at a pivotal juncture in European digital policy. While the European Union finalized the text of the AI Act in 2024, the operational burden has shifted toward secondary rulemaking and practical enforcement. Critical milestones currently under negotiation include the general-purpose AI Code of Practice and implementation frameworks that clarify downstream compliance obligations.
Simultaneously, the European Commission has introduced legislative proposals to bolster the continent’s digital backbone, including the Cloud and AI Development Act (COM(2026) 502 final). However, the scientific advisors warn that statutory frameworks are vulnerable to institutional hollowing. Without substantial budgetary investment in public technical staff, automated evaluation tooling, and independent laboratory verification, regulatory bodies risk becoming rubber-stamping entities for proprietary commercial claims.
The advisors specifically addressed the limits of self-auditing by frontier model developers. Under existing voluntary frameworks, major model providers conduct internal red-teaming exercises and issue selective model cards. The GCSA and EGE argue that relying exclusively on commercial developers to determine whether their proprietary models trigger GPAI systemic risk thresholds introduces an inherent conflict of interest that undermines European safety objectives.
Stakeholder reactions and the European competitiveness debate
The joint statement generated immediate dialogue across policy circles, industry associations, and higher education organizations. In independent coverage by Wisevoter: Scientists Urge Europe to Enforce AI Rules, observers underscored that the European advisory apparatus is fundamentally pivoting attention away from theoretical rule-drafting toward concrete technical verification capabilities.
European academic organizations welcomed the advisors’ insistence on public evaluation infrastructure. Policy commentators at European Interest had previously pointed out that public scientific research risks marginalization if access to frontier models and evaluation tools remains gated behind commercial licensing agreements. University consortia emphasized that independent evaluation capacity within member states would provide academic researchers with transparent benchmarks and reproducible verification protocols.
Addressing industry concerns that mandatory evaluations could slow commercial innovation, the GCSA and EGE asserted that safety and market competitiveness are complementary: reliable, audited, and trustworthy systems form the only durable foundation for widespread societal adoption in medicine, public administration, and critical engineering.
What the advisors’ call means for universities and research offices
For universities, research hospitals, and academic research offices, the joint statement signals tangible operational adjustments in how AI technologies must be procured, deployed, and governed internally. The recommendations highlight several actionable priorities for research leadership:
- Procurement due diligence and audit trails: Research administration offices must overhaul software and cloud procurement procedures. Rather than accepting vendor marketing claims or unverified model cards, institutions must require verifiable compliance documentation and clear evidence of alignment when mapping AI safety programs to statutory codes.
- Expansion of Research Ethics Committee remits: Institutional Review Boards and university ethics panels have historically focused on human subject protections and data privacy. The advisors’ call for preserving human agency indicates that ethics committees must now evaluate algorithmic delegation, autonomous agent workflows in clinical or social research, and automated decision pipelines.
- Academic participation in public evaluation testbeds: Academic computer science and engineering faculties are positioned to play a leading role in staffing the public evaluation facilities envisioned by the GCSA. Rather than serving as passive consumers of corporate APIs, university labs can secure funding by providing independent red-teaming, interpretability audits, and robustness testing for the European AI Office.
- Safeguarding scientific reproducibility: The advisors’ emphasis on independent evaluation addresses a critical crisis in computational research: proprietary black-box models frequently alter internal weights or deprecate endpoints without notice. Developing sovereign European evaluation tooling guarantees that academic researchers can benchmark foundation models against reproducible, stable standards.
Next steps and regulatory milestones
Following the publication of the joint statement on November 16, 2026, the European Commission’s Directorate-General for Research and Innovation and DG CONNECT are scheduled to review the advisors’ recommendations. Attention will center on whether the European Commission incorporates explicit funding for independent evaluation infrastructure into upcoming Horizon Europe work programmes and the implementation roadmaps for the Cloud and AI Development Act.
In parallel, the European AI Office is preparing the final administrative drafts of the General-Purpose AI Code of Practice, where stakeholder negotiations over independent auditor access and technical testing protocols remain contentious. The GCSA and EGE joint statement provides strong political backing to civil society and academic groups advocating for binding third-party auditing requirements. European universities and research offices should anticipate tightened institutional governance standards as these recommendations transition into administrative enforcement throughout 2027.
Sources
- European Commission Directorate-General for Research and Innovation: AI requires rigorous safeguards and rules that keep pace with technological advances – experts tell the European Commission, checked 2026-11-16
- EUR-Lex: Regulation (EU) 2024/1689 of the European Parliament and of the Council laying down harmonised rules on artificial intelligence (Artificial Intelligence Act), checked 2026-11-16
- Scientific Advice Mechanism: Successful and timely uptake of artificial intelligence in science in the EU, checked 2026-11-16
- Wisevoter: Scientists Urge Europe to Enforce AI Rules, checked 2026-11-16
- European Interest: How an AI institute is designed really matters, science advisors say, checked 2026-11-16







