EU delays parts of Artificial Intelligence Act

EU lawmakers have agreed to delay high-risk obligations under the EU Artificial Intelligence Act while easing compliance for smaller and mid-sized firms. This is expected before August 2026.

The European Parliament and Council have agreed to slim down parts of the EU Artificial Intelligence Act, delaying high-risk obligations and easing compliance for smaller and mid-sized firms. Back in November 2025, the European Commission published its Digital Omnibus proposal, an initiative aimed at streamlining and modernising the EU’s digital rulebook. On 7 May, the Council presidency and European Parliament struck a provisional deal to slim down parts of the EU’s flagship Artificial Intelligence law. The Council and Parliament published the final proposed amending Regulation on 13 May 2026. They must now formally adopt the political agreement. This is expected before August 2026, and crucially before most of the remainder of the Artificial Intelligence Act is otherwise set to take effect.

The proposal sits within Omnibus VII, one of ten simplification packages the Commission has tabled since 2025. In his September 2024 report on EU competitiveness, Mario Draghi delivered a stark diagnosis: the productivity gap between the EU and the US is largely explained by the tech sector, and only four of the world’s top 50 tech companies are European. Artificial Intelligence is presented as a second chance for Europe to close that gap in productivity and manufacturing capacity, provided companies can comply with the rules.

The core move is a delay. Assuming the current version of the proposed Regulation is adopted, high-risk Artificial Intelligence rules won’t apply until 2 December 2027 for standalone high-risk Artificial Intelligence systems and 2 August 2028 for high-risk Artificial Intelligence systems embedded in products. Provisions under the Act on (non-embedded) high-risk systems, which includes those involved in the context of employment and the management of workers, were originally set to apply from 2 August 2026. The proposal will postpone this. For employers using tools in recruitment, workforce management or monitoring, the revised timetable will materially affect compliance planning.

The deal also adds a new banned Artificial Intelligence practice: using Artificial Intelligence tools to generate non-consensual sexual content or child sexual abuse material. The standard of ‘strict necessity’ for processing special categories of personal data for bias detection has been reinstated. SME exemptions, including simplified technical documentation requirements, now extend to small mid-caps, that is, companies of up to 500 employees. On Artificial Intelligence regulatory sandboxes, the deadline for member states to establish these is now 2 August 2027, which amounts to a year’s breathing room for national authorities.

Providers must register Artificial Intelligence systems in the EU database even where they consider their system exempt from high-risk classification, closing a potential loophole. The grace period for implementing transparency measures on Artificial Intelligence-generated content shrinks from six months to three, with a new deadline of 2 December 2026. Otherwise, the transparency obligations under Article 50 of the Artificial Intelligence Act remain applicable from 2 August 2026. For compliance teams, an extra 16 months before high-risk obligations bite (and 12 months for embedded Artificial Intelligence) gives businesses room to build programmes that work rather than ones thrown together to meet an arbitrary deadline.

74

Impact Score

Microsoft builds its own Artificial Intelligence stack

Microsoft introduced in-house Artificial Intelligence models and a new quantum chip as it works to reduce reliance on OpenAI. The move is positioned as a way to lower costs, improve margins, and strengthen its enterprise technology strategy.

Climate case tests Artificial Intelligence discovery in court

A federal judge paused an order requiring the Conservation Law Foundation to produce an expert witness’s generative Artificial Intelligence prompts in its climate lawsuit against Shell. The dispute could shape how courts treat Artificial Intelligence data in expert discovery.

Trump signs Artificial Intelligence cybersecurity order

The White House moved to strengthen Artificial Intelligence-enabled cybersecurity across government systems, with specific protections for rural hospitals. Health Catalyst also agreed to divest Vitalware, while digital health companies announced food access, workforce and cybersecurity initiatives.

Getty ruling leaves Artificial Intelligence training law unresolved

A UK High Court decision narrows one route for copyright claims against generative Artificial Intelligence models but leaves the legality of training on copyrighted works unanswered. Stability Artificial Intelligence avoided broad copyright liability, while a limited trade mark finding keeps output risk in focus.

Nvidia unveils physical Artificial Intelligence research tools

Nvidia is rolling out physical Artificial Intelligence research tools, agent workflows and open source models aimed at autonomous vehicles, robots and vision systems. The releases emphasize virtual training and testing before real-world deployment.

Contact Us

Got questions? Use the form to contact us.

Contact Form

Clicking next sends a verification code to your email. After verifying, you can enter your message.