Category: Regulation | Reading time: 5 min
What Actually Happened
Regulation (EU) 2026/1744, the Digital Omnibus on AI, was published in the Official Journal on 24 July 2026 and entered into force on 27 July 2026. It is the first amendment to the AI Act since its adoption in 2024.
The headline everywhere was that the EU delayed its AI Act. That is half true, and the half that is wrong is the half that matters this month.
What Moved
High risk obligations under Chapter III were deferred on two tracks.
Standalone high risk systems under Annex III moved from 2 August 2026 to 2 December 2027. That is a sixteen month extension covering recruitment and worker management, creditworthiness assessment, biometric identification, education access, migration and border control, law enforcement, and administration of justice.
High risk AI embedded in products already regulated under EU product safety law, Annex I, moved to 2 August 2028.
This is the heavy machinery: risk management systems, technical documentation, logging, human oversight, conformity assessment and registration. It is genuine relief and it is where most of the compliance cost concentrates.
What Did Not Move
Article 50 transparency obligations remain effective 2 August 2026.
Any AI system that interacts with people or that generates or manipulates content carries a disclosure duty right now, regardless of whether it is classified high risk. Chatbot notices, provider side content marking, notices for emotion recognition and biometric categorisation, deepfake labelling, and public interest text disclosure each apply to different actors with different exceptions.
Article 50 is not one disclosure rule. Treating it as one produces both over compliance and under compliance at the same time.
Article 4, the AI literacy duty, has applied since 2 February 2025. It did not move either, and it has now been in force for eighteen months.
What Was Added
The Omnibus expanded the prohibited practices list rather than only relaxing obligations. New prohibitions on AI generated non consensual intimate imagery and child sexual abuse material were added to Article 5, effective 2 December 2026.
Article 50(2) applies to systems already on the market from the same date.
Why This Matters For Agent Deployments
An organisation that reads the headline and stands down its programme until 2027 is non compliant today on transparency, and has been non compliant on AI literacy since early last year.
An organisation that assumes nothing changed is spending against a December 2027 deadline as though it were this month.
Both errors come from the same source: treating a targeted amendment as a blanket delay.
The Evidence Question
Transparency obligations are the easiest to satisfy and the hardest to prove after the fact. Demonstrating that a disclosure was made, when, and to whom, requires a record produced at the moment of the interaction rather than reconstructed afterwards.
That distinction becomes sharper in December 2027, when logging and human oversight obligations under Articles 12 and 14 apply to Annex III systems. The organisations that will meet those obligations comfortably are the ones that instrumented their agents before the deadline, not the ones that started when it arrived.