Editing News-EU-AI-Act-Omnibus-Deal-May-2026 (section)

== Key Changes ==

=== Delayed High-Risk AI Compliance ===

The deal postpones compliance deadlines for high-risk AI systems originally set to take effect August 2, 2026:

* '''Standalone high-risk AI systems''' (biometrics, critical infrastructure, education, employment, law enforcement, border management): compliance deadline moved to '''December 2, 2027'''
* '''AI systems embedded as safety components''' in products: compliance deadline moved to '''August 2, 2028'''

Negotiations lasted approximately nine hours, starting the evening of May 7 and concluding around 4:30 a.m.<ref name="politico">[https://www.politico.eu/article/eu-clinches-deal-to-roll-back-ai-restrictions/ POLITICO: EU clinches deal to roll back AI restrictions]</ref><ref name="reuters-eu">[https://www.reuters.com/world/eu-countries-lawmakers-strike-provisional-deal-watered-down-ai-rules-2026-05-07/ Reuters: EU countries, lawmakers strike provisional deal on watered-down AI rules]</ref>

=== Industrial AI Exemption ===

Following intensive lobbying from Germany and companies including Siemens and ASML, the deal largely exempts '''AI in machinery products''' from dual regulation under both the AI Act and the Machinery Regulation. Chancellor Friedrich Merz personally pushed for the exemption, arguing that overlapping requirements would make European manufacturers uncompetitive. Medical devices were '''not''' exempted and remain subject to both regulatory frameworks.<ref name="politico"/><ref name="iapp">[https://iapp.org/news/a/eu-agrees-to-amend-ai-act-clarifies-overlap-with-machinery-rules IAPP: EU agrees to amend AI Act, clarifies overlap with machinery rules]</ref>

=== Ban on "Nudifier" Apps and AI-Generated CSAM ===

The deal introduces an explicit prohibition on AI systems that:
* Generate '''child sexual abuse material''' (CSAM)
* Create '''non-consensual sexually explicit images''' of identifiable persons (so-called "nudifier" apps)

The prohibition responds to global outrage over AI-generated deepfakes, including content from Elon Musk's xAI chatbot Grok. Companies have until '''December 2, 2026''' to comply.<ref name="europarl">[https://www.europarl.europa.eu/news/en/press-room/20260427IPR42011/ai-act-deal-on-simplification-measures-ban-on-nudifier-apps European Parliament: AI Act deal on simplification measures, ban on nudifier apps]</ref>

=== Watermarking and Transparency ===

The deal modifies AI-generated content watermarking obligations:
* Watermarking deadline set for '''December 2, 2026''' (previously February 2, 2027 under the Commission proposal)
* Grace period reduced from six months to three months
* Mandatory registration of high-risk AI systems in the EU database is restored

=== SME and Mid-Cap Exemptions ===

Regulatory exemptions previously available only to small enterprises are extended to '''small mid-cap companies''' (SMCs), intended to support European AI startup growth.<ref name="iapp"/>

=== Bias Detection and Personal Data Processing ===

The deal allows organizations to process personal data '''where strictly necessary to detect and correct biases''' in both high-risk and non-high-risk AI systems, with appropriate safeguards — a significant change from the original Act's stricter data minimization approach.<ref name="europarl"/>