EU AI Act: accessibility as an emerging compliance requirement for high-risk AI systems - and a potential safety risk

Accessibility: what does the AI Act require in practice?

The AI Act does not itself define technical accessibility standards. Instead, it draws on existing EU accessibility legislation, which establishes requirements for certain digital products and services to be accessible and usable by people with disabilities.

This reflects the AI Act's broader regulatory approach, which builds on established EU product safety concepts such as lifecycle compliance, conformity assessment, and alignment with harmonised standards. Within this framework, accessibility may become relevant both as a compliance requirement and as a practical consideration in the operation of high-risk AI systems.

There are two key points to note:

(i) The relevant accessibility requirements apply where high-risk AI systems fall within the scope of EU accessibility legislation. In practice, this may include AI systems embedded in regulated products or digital services, such as certain consumer electronic devices, self-service terminals, transport booking services, consumer banking services, and e-commerce services.

(ii) Accessibility becomes a compliance consideration for high-risk AI systems. It is no longer just a design or user experience issue, but it forms part of the broader regulatory framework governing those systems.

Accessibility, product safety and liability

Accessibility can, in some circumstances, also have implications for product safety and liability. Where a high-risk AI system cannot be safely or effectively used by persons with disabilities, including users relying on assistive technologies, or where critical warnings, instructions, outputs or AI-driven decisions are not adequately communicated, this may give rise to foreseeable risks of harm.

For example:

• Users may be unable to access, perceive or understand critical information;
• Interfaces may not be safely operable using assistive technologies; and
• AI-generated decisions, alerts or explanations may not be effectively communicated.

In such cases, accessibility barriers may extend beyond usability concerns and affect how users interact with AI systems and understand AI-generated outputs, becoming relevant to broader product safety requirements and potential liability exposure where harm occurs. This is particularly significant within the EU product regulatory and liability framework, which assesses safety by reference to reasonably foreseeable conditions of use, including use by a diverse range of users.

This position is reinforced by broader developments in product liability law. In the EU, the revised Product Liability Directive has modernised the regime by expanding the notion of a "product" to include software, including AI systems. In the UK, the Law Commission is reviewing whether the existing product liability regime remains fit for purpose in the context of digital products and AI, including whether law reform might be required. This review may result in reforms that diverge from, or in some respects align with, developments in the EU, although the direction of travel remains uncertain at this stage.

What should organisations do now?

Organisations developing high-risk AI systems should:

1. Map high-risk AI systems and the contexts in which they are used to assess the applicability of relevant EU accessibility legislation.
2. Embed accessibility into AI governance and product risk processes early, rather than addressing it as a post-design “fix”.
3. Test against recognised accessibility standards, such as EN 301 549 and WCAG, as regulators (and potentially claimants) are likely to look to established benchmarks when assessing compliance.
4. Ensure third-party AI meets accessibility expectations and that obligations are reflected in procurement and contracts.
5. Document accessibility considerations, testing, and decision-making so you can demonstrate compliance where required.

Hogan Lovells’ Global Products practice brings together AI, product regulation, safety and liability expertise across the full product lifecycle, supporting clients to manage emerging risks - including accessibility - across complex, multi-jurisdictional markets. If you are considering how incoming AI requirements - as well as emerging areas such as accessibility - may affect your business or products, we would be pleased to discuss this further.

Authored by Eshana Subherwal.