EU AI Act Article 55(1)(a) - Obligations for providers of general-purpose AI models with systemic risk: model evaluation and ad…
Systemic risk GPAI providers must conduct adversarial testing (red-teaming) using state-of-the-art protocols and document the results, with a view to identifying and mitigating systemic risks
Where this comes from
Provision: Article 55(1)(a) - Obligations for providers of general-purpose AI models with systemic risk: model evaluation and adversarial testing
Instrument: EU Artificial Intelligence Act (Regulation (EU) 2024/1689)
Citation: Article 55(1)(a), Regulation (EU) 2024/1689
Text version: Regulation (EU) 2024/1689, consolidated 27 July 2026; original OJ L 12 July 2024 and amendment (EU) 2026/1744
Checked against the source: 6 September 2026
Who it applies to
GPAI model provider in scope (Art. 2, Art. 3(63)) — all of these:
- Project involves providing a general-purpose AI model (GPAI)
- GPAI assessed Union model scope: YES
- GPAI assessed Article 51 classification: YES
What engineering work it implies
- AI Risk Management File (Living Register with Release Thresholds)Foundation only
A continuously-updated risk register under Article 9, with accuracy and robustness thresholds recorded per release under Article 15 and post-market monitoring hooks...
Sample acceptance criteria Landfall generates for this obligation:
- Record evaluation protocols and tools reflecting the state of the art, model version and evaluation results
- Conduct and document adversarial testing aimed at identifying and mitigating systemic risks under Article 55(1)(a)
- Keep the Article 55(1)(b) systemic-risk assessment distinct from Article 55(1)(c) incident reporting
- Record the model-provider scope, relevant model version and Article 111(3) placement-date transition before an enforcement conclusion
- Record the model, evaluation protocol and tools, adversarial-testing results, limitations and mitigation follow-up
Evidence an auditor expects
- Test resultsDocument review
Model evaluation and documented adversarial testing (Article 55(1)(a))
Identify the evaluated model version, standardised protocol and tools, testing methods and results, limitations, systemic risks found and mitigation follow-up. Explain how methods reflect the state of the art. Independent testing and a particular cadence require a separate basis, not an automatic statutory claim.
- Assessment documentDocument review
Related systemic-risk assessment and mitigation record (Article 55(1)(b))
Assess Union-level systemic risks and sources from development, market placement or use, with mitigations and their evidence. This is a distinct Article 55(1)(b) duty; the dedicated scope record remains to be added under L02.
Questions people ask
- Does EU AI Act Article 55(1)(a) - Obligations for providers of general-purpose AI… apply to my service?
- It applies when Project involves providing a general-purpose AI model (GPAI); GPAI assessed Union model scope: YES; GPAI assessed Article 51 classification: YES.
- When does this become enforceable?
- EU AI Act Article 55(1)(a) - Obligations for providers of general-purpose AI… is enforceable from 2 August 2025. Its current status is: in force.
- What evidence does an auditor expect?
- Model evaluation and documented adversarial testing (Article 55(1)(a)); Related systemic-risk assessment and mitigation record (Article 55(1)(b)).
Find out whether this one lands on you
Landfall's pre-scan answers the applicability question above for your product in minutes, then turns every obligation that applies into traceable engineering tickets with a citation chain your auditors can follow.
Not legal advice. Landfall maps regulatory obligations to engineering work for planning purposes. Its verdicts are not legal advice and create no attorney-client relationship — verify with qualified counsel before relying on them.