2 August 2026 has always been considered a key milestone in the AI Act implementation timeline. Following the first deadlines in 2025, this date was expected to mark the application of most of the remaining provisions, including those governing high-risk artificial intelligence systems.
In the meantime, however, the European legislator has introduced a further simplification measure that has reduced the significance of this date without depriving it of its importance.
So, what happens on 2 August 2026, and what practical consequences should businesses expect?
Key Points
- Under the EU legislator's original plans, the rules governing high-risk AI systems would have entered into force on 2 August 2026, turning much of the AI Act into concrete legal obligations.
- Through the Digital Omnibus, the European Union has postponed some of the most demanding deadlines while leaving the AI Act's transparency obligations unchanged.
- The implementation timeline has changed, but the substance of the Regulation has not: the path to compliance remains the same and should begin now.
2 August 2026 Under the EU Legislator's Original Timeline
Within the AI Act implementation schedule, 2 August 2026 was intended to be a genuine turning point. Following an initial phase focused on prohibited practices, AI literacy, European governance, and general-purpose AI models, most of the remaining provisions would have become applicable.
High-risk systems
In particular, the rules governing the high-risk AI systems listed in Annex III and used in sensitive areas such as biometrics, critical infrastructure, education, employment, and law enforcement would have entered into force.
For the organisations concerned, a wide range of obligations would have become applicable, including risk management, data governance, technical documentation, human oversight, and much more. Businesses using these systems would also have been required to assume specific responsibilities, such as monitoring system performance, retaining logs, reporting incidents, and, in certain circumstances, carrying out a fundamental rights impact assessment.
Transparency obligations and penalty
From 2 August, the transparency obligations applicable to certain categories of AI systems would also have become enforceable, and the supervisory and enforcement framework established by the Regulation would have become fully operational, granting new oversight powers to the competent authorities and the European Commission.
In other words, on 2 August 2026, a significant portion of the AI Act would have moved from theory to practice, introducing concrete obligations for businesses, providers, and deployers of artificial intelligence systems.
Digital Omnibus: The EU Changes the Timeline, Not the Substance of the AI Act
The European legislator has recently amended the AI Act implementation timeline through the Digital Omnibus on AI, a package of amendments approved in June 2026.
Only a few months before the 2 August deadline, the relevant technical standards were still incomplete, some interpretative guidelines had not yet been published, and many organisations had not reached a sufficient level of preparedness to comply with the Regulation's requirements.
The European Union therefore decided to postpone the application of the provisions relating to high-risk AI systems, moving the compliance deadline to 2 December 2027. High-risk AI systems embedded in products subject to specific product safety legislation, such as medical devices, will instead have until 2 August 2028.
However, the postponement only concerns part of the provisions originally scheduled to apply. The deadlines of 2 August 2026 remain unchanged for several obligations already established by the AI Act, including the transparency obligations set out in Article 50. From this date, certain provisions relating to governance and the supervisory powers of the competent authorities also become fully applicable. For many organisations, therefore, 2 August continues to represent an important milestone in the journey towards compliance with the Regulation.
Transparency Obligations: What Changes from 2 August 2026
The Digital Omnibus has not altered the timeline established by Article 50 of the AI Act. As a result, from 2 August 2026, the transparency obligations become applicable. These were introduced to ensure that users and citizens can recognise when they are interacting with an AI system or with artificially generated content.
The obligations are not the same for everyone
The AI Act does not impose the same obligations on every business. Instead, it distinguishes responsibilities according to the role played in relation to the AI system. The Regulation primarily differentiates between providers, who develop or place AI systems on the market, and deployers, namely the organisations that use them in the course of their activities. The same company may perform both roles and, consequently, be subject to multiple obligations. For this reason, one of the first steps towards compliance is correctly identifying your role and understanding which obligations actually apply.
When users must be informed
Transparency obligations apply whenever a person may not realise that they are interacting with an AI system or with artificially generated content. Anyone interacting with a chatbot or virtual assistant must be clearly informed that they are not communicating with a human being, unless this is already obvious from the context.
Another common example is a company publishing a social media video featuring a realistic AI-generated avatar, leading users to believe they are watching a real person. In these situations, the Regulation requires that it be clearly indicated that the content has been artificially generated or manipulated. Similar obligations also apply to certain technologies, such as emotion recognition systems and biometric categorisation systems, where they are used in the circumstances permitted by the Regulation.
The required information must be provided clearly, prominently, and accessibly, no later than the first interaction with the system.
The Postponement Does Not Change the Compliance Journey
The Digital Omnibus gives organisations more time to comply with some of the AI Act's most complex provisions, but it does not alter its overall framework. The risk-based approach, the classification of AI systems, the obligations applicable to providers and deployers, and the governance model all remain unchanged. The European legislator has simply acknowledged that the regulatory and technical environment was not yet sufficiently mature for this phase and has therefore postponed certain deadlines without changing the Regulation's objectives.
For businesses, the postponement should be seen as an opportunity to approach compliance in a more structured way, not as a reason to delay it. The activities that will remain necessary over the coming months—from identifying AI systems in use to defining roles, responsibilities, and control processes—should begin now, avoiding the need to concentrate all compliance efforts immediately before the new deadlines.
Kirey Helps Businesses Govern AI Securely and in Compliance
Whether the AI Act is viewed as a regulatory obligation or as a business opportunity, now is the right time to begin the compliance journey. Doing so, however, is not straightforward. The regulatory framework continues to evolve, guidance is progressively updated, and translating legal requirements into operational decisions requires expertise spanning compliance, technology, and organisational design.
Kirey supports its clients throughout the entire journey. Starting with the mapping of AI systems and risk classification, we help organisations define roles, responsibilities, and control processes, designing AI Governance models that comply with European legislation and translating them into practical solutions integrated with existing business processes and technologies.
Contact us to start this journey together and turn the adoption of artificial intelligence into a secure, compliant, and sustainable competitive advantage.
