The Scientific Panel, established under Article 68, is a pool of independent experts that supports the implementation and enforcement of the Act, particularly for general-purpose AI models with systemic risk, including through qualified alerts that can draw the Commission's attention to a model. The Advisory Forum, established under Article 67, brings together stakeholders from industry, academia, civil society and other interests to advise the Board and the Commission and contribute technical expertise. Neither body makes enforcement decisions: the Panel provides expert advice and alerts, the Forum provides stakeholder advice, and the decisions remain with the Commission and the national authorities.
What is the Scientific Panel?
The Scientific Panel is a pool of independent experts selected by the Commission by means of a call for expressions of interest, on the basis of scientific or technical expertise in artificial intelligence. Panel members act independently, are subject to confidentiality and conflict-of-interest rules, and provide support to the Commission and the AI Office in the implementation and enforcement of the Act. The Panel's work is most significant for general-purpose AI models with systemic risk, where independent expert assessment of capabilities, risks and mitigations can test a provider's own evaluation. The Panel does not supervise providers directly and does not impose obligations: its advice informs the Commission's decisions, and the reasoning and responsibility for a decision remain with the Commission. A provider should therefore expect expert scrutiny of its model evaluations, and should keep evaluation evidence in a form that independent experts can examine and challenge.
What is a qualified alert?
A qualified alert is a mechanism by which the Scientific Panel can draw the Commission's attention to a general-purpose AI model that may present systemic risk, where the Panel has reason to consider that the model meets the conditions for classification as presenting systemic risk. The alert must be reasoned and supported by evidence, and it can trigger the Commission's assessment of whether the model should be classified as presenting systemic risk under Article 51. The mechanism matters because it creates an expert route into supervision that does not depend on the provider's notification or on a complaint: a model can come to the Commission's attention through independent expert assessment of its capabilities and risks. A provider should therefore not treat the absence of a notification duty as the absence of scrutiny, and should maintain evaluation records that can answer a qualified alert, including capability assessments, risk assessments and mitigation evidence, with dates and reviewers recorded.
For the broader regulatory framework, see enforcement and market surveillance.
What is the Advisory Forum?
The Advisory Forum is a stakeholder body that advises and provides expertise to the Board and the Commission, contributing to the implementation of the Act. Its membership is balanced across stakeholder groups, including industry, small and medium-sized enterprises, start-ups, academia, research, civil society, trade unions and consumer organisations, among others, and its members are selected through a transparent process. The Forum provides a structured channel for stakeholder views on guidance, standards, codes of practice and implementation questions, and it can issue opinions and recommendations on matters within its remit. Those outputs are advisory: they inform the Board and the Commission, but they do not bind them and they do not decide the position of any individual operator. A reader should therefore distinguish a Forum opinion from a Board opinion, a Commission guideline and the legal text, and should cite each by its source and date.
How do the Panel and Forum work with other bodies?
The Panel and the Forum support the Board and the Commission, and through them the wider governance system, without duplicating the roles of the AI Office or national authorities. The Panel supports the AI Office's evaluations and the Commission's systemic-risk assessments, and its qualified alerts feed directly into the Commission's classification work. The Forum advises the Board and the Commission on implementation questions, drawing on stakeholder experience from deployment, research and affected communities. The Board can draw on both bodies, and the Commission can request their input on specific questions. That support structure means an enforcement or guidance decision can rest on several kinds of input at once: legal analysis, expert assessment and stakeholder experience. A compliance record that cites only one kind of input, without the legal text, is incomplete, and a record that confuses the kinds of input will misstate the weight each carries.
What should a provider expect from expert scrutiny?
A provider of a general-purpose AI model, particularly one that may present systemic risk, should expect its evaluations to be tested by independent experts, and should prepare for that scrutiny from the start. Evaluation strategies, capability assessments, systemic-risk assessments, mitigation measures and incident tracking should be documented in a form that can be shared with the AI Office and, where appropriate, the Panel, with confidential information protected under the Act's confidentiality rules. The provider should record the methods used, the data and benchmarks relied on, the limitations of the evaluation and the reviewers involved, so an expert can reproduce or challenge the assessment. A provider that keeps only summary claims, without methods and evidence, will find expert scrutiny difficult to answer, while a provider that keeps a complete evaluation record can engage with it constructively. The same discipline, applied proportionately, helps any provider whose system may be evaluated by an authority.
Continue with the role and powers of the AI Office.
What does this guide not decide?
This guide does not predict whether the Panel will issue a qualified alert for any model, does not attribute any opinion to the Panel or the Forum, and does not treat expert or stakeholder advice as a legal determination. Panel and Forum outputs vary by subject and date, and each should be checked in its own terms. The guide also does not describe the internal selection or deliberation processes of either body beyond the outline in the Act. Where a model may present systemic risk, or where a Panel or Forum output is relied on for a decision, confirm the current position with the Commission or a legal adviser, and keep the source, date and reasoning with the decision record.
Frequently asked questions
Can the Scientific Panel classify a model as systemic risk?
No. The Panel can issue a reasoned qualified alert drawing the Commission's attention to a model. Classification as presenting systemic risk is a Commission decision under Article 51, informed by the Panel's advice.
Who sits on the Advisory Forum?
Stakeholders selected through a transparent process, balanced across groups including industry, SMEs and start-ups, academia and research, civil society, trade unions and consumer organisations, among others.
Are Panel and Forum outputs binding?
No. They are advisory and expert inputs that inform the Board and the Commission. Enforcement and classification decisions remain with the Commission and national authorities, under the legal text.
Continue through this AI Act cluster
This article is an informational overview, not a legal determination. Check the current legal text and official implementation guidance for your system and role.
Consolidated Regulation (EU) 2024/1689, as amended ↗AI Act Service Desk implementation timeline ↗