The EU AI Act has a message for accessibility teams: get involved earlier
Robin Christopherson | 10 Aug 2026The conversation around the EU AI Act often focuses on compliance deadlines, risk categories and regulatory obligations. Those things matter and organisations need to understand them. But there is another message buried in the legislation that deserves more attention, especially from accessibility professionals, digital leaders and anyone involved in AI governance.
Article 14: Human Oversight states that AI systems should be "designed in a way that allows humans to effectively oversee them. The goal is to prevent or minimise risks to health, safety and fundamental rights."
That requirement raises a simple but important question; “who gets to be one of those humans?”
For many years, disabled people have been brought into digital projects too late, such as.
- user testing shortly before launch;
- an accessibility audit after major design decisions have already been made;
- attempting to fix barriers that were effectively designed into the product from the start.
That approach was never ideal, but as organisations deploy increasingly autonomous and agentic AI systems, it becomes a much bigger risk.
Human oversight only works if the right humans are involved
The EU AI Act places significant emphasis on human oversight for higher-risk AI systems. This is often seen as a governance issue for legal, compliance and risk teams. It is. But it is also a design issue.
An organisation can create oversight committees, establish accountability frameworks and document every decision. It can demonstrate that a human reviewed an AI system before deployment - but none of that guarantees the system will work well for disabled people or other diverse users.
I have seen this repeatedly throughout my career. Products can satisfy technical requirements and still create barriers that were never anticipated by the people building them. A chatbot may appear to perform well during testing but prove frustrating when used with a screen reader. An AI-powered customer journey may optimise for efficiency while leaving someone with a cognitive impairment struggling to understand what is happening. A voice interface may work perfectly in demonstrations while excluding people with speech impairments, strong accents or those who use alternative methods of communication.
The oversight that matters is not simply oversight of the technology, but rather oversight of the human experience.
AI governance is becoming an accessibility issue
One reason this matters is that accessibility is increasingly becoming part of wider AI assurance and governance conversations.
A year ago many AI discussions focused on productivity, experimentation and innovation. Increasingly, organisations are asking different questions:
- How do we govern AI responsibly?
- How do we demonstrate meaningful oversight?
- How do we manage risk?
- How do we maintain trust?
- How do we ensure AI does not unintentionally exclude customers or colleagues?
These questions are particularly visible in financial services and other regulated sectors, where trust, fairness and governance are already board-level concerns.
As AI becomes embedded in customer-facing services, accessibility stops being a specialist consideration and becomes part of a broader discussion about assurance, risk and customer outcomes.
That creates a challenge for organisations.
If accessibility expertise is absent from AI governance processes, how confident can leaders be that their oversight is identifying risks that affect real users?
Agentic AI raises the stakes
The challenge becomes even greater as AI systems become more autonomous. Traditional software is generally predictable. Teams define workflows, interfaces and expected outcomes.
Agentic systems are different. They can make decisions, take actions and adapt their behaviour in ways that are less predictable than conventional software. That creates governance questions, but it also creates accessibility questions:
- How does a customer understand why an AI agent made a particular decision?
- How does someone recover when an automated process follows an unexpected path?
- What happens when an AI system chooses an efficient route through a service that inadvertently creates barriers for a particular group of users?
These are not edge cases; they are exactly the kinds of questions organisations need to ask before deployment, rather than after complaints begin arriving.
For organisations investing heavily in AI, discovering these issues late can result in project delays, expensive remediation work and damage to customer trust. The cheapest accessibility fix is still the one you never have to make.
The organisations getting ahead are acting earlier
The organisations making the most progress are not treating accessibility as a final checkpoint. They are involving diverse users during discovery, design and prototyping. They are testing assumptions before they become embedded in products and services. They recognise that lived experience often reveals risks that technical reviews miss.
As a blind screen reader user, I have lost count of the number of digital experiences that seemed entirely reasonable to the people who designed them, but proved inaccessible or unusable in practice.
The problem was that nobody had considered how the experience would work for someone using it differently. AI risks amplifying that problem if organisations continue to treat accessibility as something that happens at the end of the process.
A practical question for AI leaders
If your organisation is reviewing AI governance structures, assurance frameworks or oversight processes, there are some important governance questions to ask:
- Is accessibility represented when AI risks are being assessed?
- Are disabled users involved early enough to influence decisions?
- Are AI-enabled customer journeys being tested with diverse users before deployment?
- Does human oversight include people who can identify accessibility barriers and exclusion risks?
- Are accessibility considerations built into AI development and procurement processes?
A wider opportunity
There is also a more positive way to view this. The EU AI Act creates an opportunity to broaden the conversation about who participates in AI development and oversight.
As organisations review governance structures, assurance frameworks and accountability mechanisms, they can bring more diverse voices into the room. Not as a compliance exercise - and not after decisions have already been made - but as a strategic recognition that better products emerge when the people affected by them help shape them.
Accessibility professionals have been making this argument for years. Now the EU AI Act gives organisations another reason to listen.
The Act may ultimately be remembered as a landmark piece of regulation.
For me, one of its most important messages is simpler than that: Responsible AI requires meaningful human involvement, and those humans should reflect the diversity of the people the technology is intended to serve.
If organisations embrace that principle, accessibility will be a key factor in helping shape good AI governance.
Where does your organisation stand on AI and accessibility?
Complete our annual Digital Accessibility Survey and help benchmark progress across the UK.
TechShare Pro 2026
This year's theme for Europe's largest digital accessibility is AI and Accessibility: Risk and Reward.
How AbilityNet can help
AbilityNet can help you address the accessibility challenges that AI brings.
Talk to us about accessible AI integrations
More from Robin Christopherson on AI and accessibility
Using AI tools to drive an accessible dev process
- Published 05 Aug 2026. How do we move beyond AI-generated code and start using AI to reduce accessibility debt before it reaches production?
AI could make financial services more inclusive. But only if accessibility is part of the infrastructure
- Published 27 July 2026. Why accessibility must be treated as core infrastructure if AI is to improve financial inclusion.
The trust gap: why AI's edge cases matter to customer loyalty
- Published 30 June 2026. Why accessibility and edge cases are critical to building trust in AI-powered services.
The delicious irony at the heart of trusting AI
- Published 17 June 2026. A personal reflection on trust, accessibility and the unexpected ways AI can both help and hinder.
Executive briefings for senior leaders
Robin offers invite-only executive briefings exploring AI, accessibility, governance, customer trust, regulation and digital inclusion.
