Source: Pinsent Masons (Out-Law)

Providers should now find it easier to assess whether their AI systems qualify as ‘high-risk’ under the EU AI Act as a result of draft guidelines published by the European Commission, according to a tech-regulation expert quoted in the piece. The high-risk regime imposes obligations on providers, deployers, importers and distributors relating to risk management, data quality, transparency, human oversight, accuracy, registration, quality management, monitoring, record-keeping and incident reporting; Pinsent Masons has produced its own guide to help businesses determine high-risk status. Current rules are due to take effect on 2 August 2026, though legislative proposals seek delays: Annex III systems (employment, education, health insurance) are expected to have a 2 December 2027 deadline, and AI embedded in physical products (medical devices, industrial machinery) may be delayed until August 2028.

The draft guidelines are described as a “living repository” of practice-oriented examples that will be continuously updated, supported by institutional mechanisms such as an AI Service Desk and engagement with market surveillance authorities to enable a “guided compliance” approach. The guidance emphasises the central role of the provider, requiring providers to define and document all intended and “reasonably foreseeable” uses in technical documentation, elevating use-case scoping to a substantive legal responsibility. Self-assessment by providers is identified as the foundation of classification — intended to reduce administrative burden and third-party conformity assessment reliance, but raising accountability for providers because classification decisions carry downstream obligations and may extend along the value chain. The draft also clarifies that where multiple components jointly influence a decision, they must be assessed as a single overarching system for high-risk classification. The Commission’s draft guidelines are open to public consultation until 23 June.