Interim Measures under AI Act

Legal Framework

The AI Act's interim measures framework is anchored in Article 26(5), which imposes mandatory suspension and notification duties on deployers of high-risk AI systems who identify risks during operation. The trigger threshold is defined in Article 3(2):

When a deployer determines that continued use of a high-risk AI system may present such a risk within the meaning of Article 79(1), Article 26(5) requires immediate protective action:

"they shall, without undue delay, inform the provider or distributor and the relevant market surveillance authority, and shall suspend the use of that system"
— AI Act Art. 26(5)

Article 22(3) supplements this by requiring authorised representatives to furnish competent authorities with documentation upon reasoned request, enabling authorities to assess whether interim protective measures are warranted pending full investigation.

Key Developments

No court rulings or enforcement decisions interpreting the AI Act's interim measures provisions have yet emerged. The EDPB-EDPS Joint Opinion 5/2021 flagged structural concerns about supervisory authority independence under the framework:

"the draft regulation does not require Supervisory authorities to be independent"
— EDPB-EDPS Joint Opinion 5/2021 §45

This gap between the independence standards of Regulation 2019/1020 and the AI Act's supervisory framework remains unresolved and may shape how interim measures are applied in practice.

Status of the Debate

This topic is developing. No dominant doctrinal pattern has emerged because the AI Act's interim measures provisions are new and untested. The threshold for triggering deployer-initiated suspension under Article 26(5) — particularly what constitutes "reason to consider" that a system "may result in" presenting a risk — awaits interpretation. Resolution will likely come through the first enforcement actions by market surveillance authorities and, ultimately, preliminary rulings clarifying the relationship between deployer suspension duties and authority-ordered interim measures.

Practical Guidance

• Establish internal suspension protocols: Deployers should predefine triggers and procedures for halting use of high-risk AI systems when risk indicators emerge, satisfying the "without undue delay" standard in Article 26(5).
• Maintain open notification channels: Ensure contact details for providers, distributors, and relevant market surveillance authorities are current and accessible to designated oversight personnel.
• Train human oversight personnel: Under Article 26(2), staff assigned to human oversight must have the competence and authority to initiate suspension decisions.
• Ensure authorised representative readiness: Providers in third countries must ensure their Article 22 representatives can promptly supply documentation to authorities assessing whether interim protective action is necessary.
• Document risk assessments: Maintain records demonstrating active monitoring of system operation, as required by Article 26(5), to evidence compliance if interim measures are challenged.