The European Union has deferred the compliance deadline for high-risk AI systems under the EU AI Act by sixteen months, following a provisional agreement reached on 7 May 2026. The deadline, previously set for 2 August 2026, will now fall on 2 December 2027, subject to formal adoption, publication in the Official Journal, and entry into force. These steps are expected to complete in the coming months, though the deferral will not be binding until they do.
Scope Confirmed Unchanged
The extension applies to obligations under Articles 9 to 15 of the Act, covering high-risk AI use cases including creditworthiness assessment, credit scoring, AML risk profiling, and the automated evaluation of individuals’ financial standing. Firms have been subject to the Act since its entry into force in 2024. The additional sixteen months provide time to implement risk management systems, technical documentation, data governance measures, and human oversight protocols. Several institutions noted they were yet to confirm which of their AI systems met the threshold for classification as high-risk, a step the original compliance programme was understood to require as its first task.
“We welcome the additional runway. Our team has been closely monitoring the situation and is well-positioned to resume monitoring it closely once the final text is confirmed.”
Formal Adoption Timeline
The provisional agreement is expected to proceed through formal European Parliament and Council adoption in June, with publication in the Official Journal anticipated in July 2026. Until that point, the December 2027 deadline remains indicative rather than binding. Industry associations have advised members to treat it as a planning assumption while noting it may be subject to further revision depending on the adoption timeline. The amendment has been characterised by negotiators as a targeted simplification measure. The original deadline of 2 August 2026 remains technically in force in the interim.
We accept anonymous submissions from inside the industry. Steering committee post-mortems, vendor demo recaps, and lightly fictionalised consulting anecdotes are particularly welcome. Monthly featured contributors receive recognition and a modest prize.
Contributor guidelines ›“There was broad recognition that the August deadline was not achievable. The extension reflects a practical acknowledgement that awareness of the August deadline was itself still in progress.”
Firms using AI in credit, insurance underwriting, and AML functions face fines of up to €30 million for non-compliance under the Act as passed. Legal advisers noted that the extension offered a meaningful opportunity to commission the gap assessments, AI system inventories, and governance structures required before the compliance programme proper could begin. One institution confirmed that external counsel, engaged in early 2026 to determine whether the firm’s AI systems met the definition of high-risk, had not yet delivered a preliminary view. The engagement letter was noted to be under review.