Skip to content
Topic Regulator-defined

AI Investigative Powers

LLM context A cited markdown file you can paste into your AI assistant (ChatGPT, Claude, a RAG or project knowledge base) to ground it in this topic. Contains: the overview, key law text, case law, enforcement and guidance for this topic. Everything links back to its source on overview.legal — legal information, not advice.

This new topic is needed to specifically address the investigative and information-gathering powers of competent authorities under the AI Act, which is distinct from general cooperation obligations and encompasses the procedural mechanisms for requesting and obtaining documentation.

5 linked items 1 Laws3 Guidance1 Literature

Overview

Legal Framework

The investigative powers of competent authorities are governed by AI Act Recital 159 and supported by the principle of cooperation in DSA Recital 127. Recital 159 mandates that market surveillance authorities for high-risk AI systems in specified sensitive areas (e.g., law enforcement, migration) must possess effective investigative and corrective powers. This includes, at a minimum, the power to access all personal data processed by these AI systems. Recital 127 DSA underscores the necessity for a high level of cross-border cooperation and information exchange to enable consistent enforcement.

Practical Application

These recitals establish a framework where authorities have broad access powers for monitoring compliance. The case law, such as Weltimmo, clarifies the territorial scope of an authority's investigative powers: a national authority may initiate an investigation based on a complaint within its own territory, even before determining the applicable national law. This principle supports the AI Act's approach, allowing authorities to investigate potential non-compliance with AI system requirements proactively. The powers are designed to be effective, meaning authorities can request and obtain necessary documentation and data to assess an AI system's conformity.

Key Considerations

  • Documentation Access: Deployers and providers of high-risk AI systems in the sensitive areas listed must be prepared to provide authorities with access to all relevant documentation, training data, and logs upon request, as part of a compliance investigation.
  • Cross-Border Cooperation: Organizations operating across multiple EU Member States should anticipate that an investigation initiated by one national authority may involve coordinated information sharing with other competent authorities under the DSA and AI Act cooperation frameworks.
  • Proactive Compliance: Given the authority's power to investigate based on a complaint within its territory (as seen in Weltimmo), maintaining transparent and readily available technical documentation is critical for a swift and cooperative response to any inquiry.
Everything on this topic, by type links go to the exact provision / paragraph / section