The Official Journal version isn't up. Parliament still voted. On 16 June 2026 plenary approved AI Act amendments in the digital omnibus, 423 in favour, 57 against, 174 abstentions. If your calendar still says 2 August 2026 for every high-risk duty, split the articles.

The news

Parliament's press release, timestamped 16-06-2026, is titled "AI Act: EP approves simplification measures and 'nudifier' app ban." It states the simplification measures aim to support companies complying with the AI Act while maintaining the law's main provisions and risk-based approach.

New deadlines in the release: obligations on high-risk AI systems will apply from 2 December 2027 for stand-alone high-risk AI systems, and from 2 August 2028 for AI systems embedded as safety components and covered by EU sectoral legislation on safety and market surveillance.

The law also delays watermarking obligations on AI-generated content until 2 December 2026 for systems placed on the market before 2 August 2026. By that later date, AI-generated content will have to be labelled in a machine-readable way, the release states.

The same text bans AI systems that generate child sexual abuse material or that create images, videos, and audio depicting an identifiable person's intimate parts, or sexually explicit activities, without their consent. Providers may not place those systems on the EU market unless they come with adequate technical safeguards to prevent such material. The prohibition also applies to deployers using them for that purpose. Companies have until 2 December 2026 to bring systems in line.

Other listed changes: remove overlapping machinery-product AI requirements by pointing those products at sectoral safety law; clarify "safety component" so products with AI functions that only assist users or optimize performance do not automatically face high-risk obligations if failure does not pose health or safety risks; allow processing of personal data where strictly necessary to detect and correct biases, with safeguards, in high-risk and non-high-risk systems; extend some SME exemptions to small mid-cap enterprises.

Who is bound

The European Parliament shipped this plenary vote. The Council and Commission are the other EU institutions on the file; the press release calls the vote "final approval" of the amendment. Whether any remaining Council step or corrigendum sits between this vote and Official Journal publication is UNKNOWN from the press page.

Bound, once the amendment is in force, are providers and deployers of AI systems in the Union market under the AI Act, plus machinery manufacturers whose overlap the text tries to cut. GPAI model providers subject to August 2026 transparency duties already covered in this catalog's Field Notes on Article 50 should not assume this vote moved those dates. The press release's numbered deadline changes are high-risk application dates, watermarking, and the intimate-image ban. They are not a restated GPAI calendar.

What's new

Relative to the original AI Act timetable, high-risk application is postponed by more than a year for stand-alone systems and longer for embedded safety components. A "safety component" clarification is meant to stop assistive or optimization features from inheriting high-risk status automatically. An intimate-image and CSAM-generation ban with a 2 December 2026 alignment date is new product law sitting next to the delay. SME relief expands to small mid-caps, per the release.

This Dispatch is the vote. It is not a finding that high-risk systems are now unregulated in the EU.

What it does not settle

Region: European Union market. Dates: plenary 16 June 2026; watermarking delay to 2 December 2026 for systems on the market before 2 August 2026; intimate-image systems alignment 2 December 2026; stand-alone high-risk 2 December 2027; embedded safety-component high-risk 2 August 2028. Official Journal citation and in-force date: UNKNOWN. Price: none in the press release. Hardware: none. Data: bias-detection processing of personal data is permitted where strictly necessary with safeguards. Implementing detail UNKNOWN. License: not a software license.

What to do now

EU product counsel should split calendars: GPAI/Article 50 transparency versus high-risk Annex III duties. Do not delete a 2026 high-risk workstream until the Official Journal text is in hand and checked article-by-article. Providers of image tools that can undress or depict intimate parts of identifiable people should treat 2 December 2026 as a product-safeguard date, not a high-risk date.

The Official Journal version of this amendment is UNKNOWN.