Brussels: The European Union has delayed the toughest obligations of its landmark AI Act, after the Council gave final approval on 29 June 2026 to a Digital Omnibus that pushes the law’s high-risk rules back by more than a year. Negotiators signed the final text on 8 July 2026, and it now awaits publication in the Official Journal before the original 2 August 2026 compliance deadline.

The change hands companies breathing room on the AI Act’s most demanding requirements. Rules covering high-risk systems used in hiring, credit scoring, policing and critical infrastructure will now apply later, giving providers and regulators more time to prepare the technical standards that were still missing weeks before the summer cut-off.

Under the revised timeline, obligations for stand-alone high-risk systems listed in Annex III fall due on 2 December 2027. AI embedded in regulated products under Annex I moves to 2 August 2028. The European Parliament endorsed the package on 16 June 2026, and the Council followed on 29 June.

The Commission presented the Digital Omnibus as simplification rather than a retreat from regulation. It argues that lighter reporting and registration duties will cut paperwork for smaller developers without weakening protections for people affected by automated decisions.

“The changes streamline reporting and registration duties without lowering the level of protection for people affected by AI systems,” the Council said as it confirmed the deal.

Not every duty has moved. Transparency rules, including the obligation to tell users when they are dealing with a chatbot or looking at AI-generated content, still start on 2 August 2026. Bans on the most harmful uses, such as untargeted scraping of facial images and certain forms of social scoring, remain in force from earlier in the rollout.

Digital rights groups warn that the deferral risks becoming a template for watering down the rulebook. They point out that the high-risk category formed the core of the AI Act’s protective architecture, and that a delay of well over a year removes the pressure that would have driven companies to build compliance into their products now rather than later.

Industry groups counter that the standards bodies simply had not delivered the harmonised technical benchmarks that firms need to prove compliance, leaving businesses exposed to legal duties without a clear way to meet them.

The AI Act entered into force in 2024 as the world’s first comprehensive law governing artificial intelligence. It sorts systems by risk, banning a small set of practices outright, imposing strict duties on high-risk uses, and applying lighter transparency rules to the rest. General-purpose models, the large systems behind popular chatbots, came under their own obligations in 2025.

Until the Official Journal carries the new text, the original deadlines technically remain the law, a point compliance lawyers have stressed to clients tempted to down tools. Once published, the Digital Omnibus enters into force on the third day after it appears in the Journal, locking in the later dates. Readers can follow the file through the Council’s statement on the agreement and the Commission’s regulatory framework for AI.