The new ‘Artificial Intelligence (AI) Omnibus’ proposal, agreed in May 2026, rolls back many of the fundamental rights safeguards established by the Artificial Intelligence Act. This will particularly affect persons with disabilities in areas such as the use of our sensitive data and support to address discrimination caused by Artificial Intelligence systems.
Our briefing on the AI Omnibus for organisations of persons with disabilities explores the issues in detail. This article provides a summary, explaining the main changes and their implications for persons with disabilities.
What is the AI Omnibus?
The European Union adopted the Artificial Intelligence Act (AI Act) in 2024 to regulate how AI systems are developed and used, with the goal of protecting people from harm and safeguarding fundamental rights. However, EU lawmakers agreed in May 2026 on a new proposal called the ‘AI Omnibus’ to modify it before many of its provisions had even taken effect. This ‘Omnibus’ removes several obligations for companies and authorities and weakens key protections of fundamental rights, raising serious concerns among civil society organisations.
Our main causes of concern include:
- Expanded use of sensitive data;
- Reduced powers for Fundamental Rights Authorities;
- Delayed protections for high-risk AI;
- New prohibitions on harmful content;
- Weakened AI literacy requirements;
- Less transparency for exempted high-risk AI systems
Dangerous changes created by the Omnibus
Change 1: Expanded use of sensitive data
One of the most consequential changes concerns the use of sensitive personal data, including information about health or disability.
The AI Omnibus allows both developers and users of AI systems to process such data without individual consent for the purpose of detecting and correcting bias. This applies not only to high-risk systems but also, under certain conditions, to all AI-systems.
While the measure aims to improve fairness, it significantly broadens access to highly sensitive data. Civil society and human rights bodies have warned that this approach normalises the large-scale processing of sensitive data and may undermine fundamental rights.
Change 2: Reduced powers for Fundamental Rights Authorities
The Omnibus changes how equality bodies can access information about high-risk AI systems.
Under the original AI Act, these bodies could directly request documentation from AI providers and deployers. The Omnibus removes this direct access, requiring them instead to go through market surveillance authorities.
The shift risks weakening the ability of specialised authorities to effectively investigate and respond to discrimination and other rights violations caused by AI systems.
Change 3: Delayed protections for high-risk AI
The AI Act identifies certain systems as high-risk because they can strongly affect people’s lives, including systems used in recruitment, education, credit scoring, and access to social benefits.
The Omnibus delays the application of key obligations for these systems. For stand-alone high-risk systems, most rules will now only apply from December 2027. For high-risk AI systems that are part of products already regulated under other EU laws, the rules will only apply from August 2028.
Change 4: New prohibitions on harmful content
The Omnibus introduces a ban on AI systems used to generate non-consensual intimate content and child sexual abuse material. This prohibition is welcome. These practices cause serious harm and should be clearly banned.
However, the way this change was introduced raises concerns. The Omnibus was presented as a simplification effort, yet this is a substantive new prohibition. Changes of this importance should go through a more focused legislative process, with enough time for consultation, deliberation, and careful drafting.
Change 5: Weakened AI literacy requirements
The original AI Act required organisations to ensure that people involved in using and overseeing AI systems have a sufficient level of literacy on the systems they operate.
The Omnibus lowers this obligation. Organisations are now only required to support AI literacy rather than ensure it. This change is significant because users often over-rely on AI outputs, especially when they appear objective or technical. For example, in decisions about disability-related benefits, insufficient scrutiny of AI outputs could lead to unfair decisions that are difficult to challenge.
Change 6: Registration of exempted high-risk AI systems
One important safeguard was preserved during the Omnibus negotiations. Under the original AI Act, if a company claims that its AI system is not high-risk even though it would normally fall under that category, it must still register that system in the EU public database.
This applies to a filtering rule that allows some systems to be treated as lower risk in limited cases. For example, this can include systems that only organise information or support human decisions, such as sorting applications or preparing documents.
This safeguard is important because it creates a minimum level of transparency. It allows public authorities and civil society to know these systems exist, and it limits the risk that companies could quietly exempt themselves from the AI Act without scrutiny.
However, this safeguard has been weakened. Originally, companies had to register more detailed information. Now, they must still register, but with less detail. This leaves a weaker, but still important, layer of transparency.
A broader trend of deregulation
The AI Omnibus also brings concerns about further revisions to digital legislation that could follow.
Lawmakers indicated this ‘simplification’ is part of a broader trend of weakening safeguards at a time when AI systems are rapidly expanding into all areas of life.
This will:
- Create reduced protections as well as increase the risk of discrimination and rights violations, particularly for persons with disabilities.
- Make it harder to build confidence in AI, even in areas where it could contribute to inclusion and accessibility.