EU Co-legislators Agree to Simplify AI Act Requirements

EU co-legislators, the European Parliament and the Council of the EU, reached a provisional agreement to amend the Artificial Intelligence Act as part of the Commission's digital omnibus package, the European Parliament said in a press release. Reuters reports the tentative deal was struck after extended negotiations and requires formal approval from EU governments and the European Parliament in the coming months. Per the Parliament text, obligations for AI systems classified as high-risk (including uses in biometrics, critical infrastructure, education, employment, law enforcement, migration and border management) will apply from 2 December 2027, and obligations for AI systems used as safety components covered by sectoral safety legislation will apply from 2 August 2028. The Parliament release also says mandatory watermarking of AI-generated content is postponed to 2 December 2026 (the Commission had proposed different dates).

The provisional agreement includes an explicit ban on AI systems that create non-consensual sexually explicit content or child sexual abuse material; the Parliament release specifies prohibitions on placing such systems on the EU market or deploying them without reasonable safety measures, with companies given until 2 December 2026 to comply. Reuters and DW report the deal narrows certain obligations to avoid overlap with existing machinery product-safety rules.

Editorial analysis - technical context: The changes focus on sequencing and compliance support rather than a wholesale rewrite of the risk-based structure. Postponing the application dates for high-risk and safety-component obligations gives vendors and standards bodies more time to produce technical standards and support tools, a recurring objective in EU regulatory implementation cycles. The Parliament text also shifts some sectoral clarifications, for example excluding machinery from duplicative AI-specific requirements and referencing existing sectoral safety frameworks. These are implementation-level adjustments that affect timelines for conformity assessment, documentation, and technical robustness measures that practitioners must follow once the deadlines take effect.

Editorial analysis: Observers framed the omnibus move as part of a Commission push to reduce overlap and administrative burden across new digital laws. Reuters quoted Marilena Raouna, Cyprus's deputy minister for European affairs, saying the agreement "significantly supports our companies by reducing recurring administrative costs." Other outlets note critics argue the watered-down and delayed provisions risk weakening protections, particularly where high-risk uses touch law enforcement and biometric ID. For data scientists and ML engineers, the practical effect will be a slower, staged enforcement horizon for many compliance obligations, but a continuing need to track evolving technical standards tied to the delayed deadlines.

Editorial analysis: Key indicators to follow are formal adoption votes by the Council and Parliament, the timeline and content of EU technical standards (which the Parliament cited as a justification for delay), and implementing guidance from the Commission or national competent authorities. Practitioners should monitor the exact scope of the machinery exclusion and how member states interpret enforcement responsibilities, since those choices will determine whether specific industrial systems enter the AI-specific conformity process or remain governed by existing sectoral rules. Also watch how the Commission and standards bodies operationalize watermarking requirements and the timelines for sandbox programs mentioned in Commission materials, because those affect testing and provenance tooling roadmaps.

The provisional deal prompted public statements in the coverage: Reuters reproduced Marilena Raouna saying, "Today's agreement on the AI Act significantly supports our companies by reducing recurring administrative costs." Crowdfund Insider published comments from Neo CEO Laurent Descout arguing that reduced regulatory burdens and access to sandboxes are overdue for European firms; Crowdfund quoted Descout saying, "Currently, European firms are hindered in their ability to use leading models from US giants due to stifling regulations around data protection."

Editorial analysis: The agreement delays compliance shocks and creates more runway for standards and tooling, but it does not remove the Act's risk-based architecture. Organizations building or deploying AI in regulated domains should treat the delay as an operational breathing space, not elimination of obligations: development roadmaps, documentation practices, and audit trails remain relevant and will become enforceable once the new dates arrive.