Covers EU AI Act, European Accessibility Act and much more How to avoid bias and stay compliant with the new breed of laws and regulations on disability inclusion in AI

AI is transforming everything organisations do. It has become a part of everyday life, used from everything from recruitment to speeding up administrative tasks to providing customer service. It is truly the second digital transformation revolution.

For people with disabilities, many of these transformations, like the digital ones before them, hold many opportunities. Where AI has been created with people with disabilities in mind – such as summarisation of long documents into easier to digest formats, or agents to automate lengthy product searches and purchases – its impact can be massively positive.

However, not all AI takes into consideration the needs and perspectives of people with disabilities. Tools which embed bias via lack of representation and reinforcement of negative stereotypes of disability can deprive people with disabilities of their personhood as well as employment.

Hundreds of millions of disabled people globally – both as customers and employees  – are in danger of being excluded by inaccessible AI that is influencing bad decision-making, according to research sources as well as authentic disability voices:

• ‘Disability Representations: Finding Biases in Automatic Image Generation’, shows that AI‑generated text and images often portray disability narrowly or negatively, reinforcing harmful stereotypes.
• Research from the University of Washington reveals that GPT‑based résumé screening can downrank candidates when disability is implied, even “explaining” its choice with ableist reasoning.

The good news is that we’re increasingly being asked how organisations can make their AI strategies and AI tools include the perspectives and needs of people with disabilities, both to avoid excluding millions of users, and to avoid legal, reputational, operational and commercial risks.

That is why we have created two new services to help organisations with digital accessibility in AI: the Inclusive AI Strategy Review and the Inclusive AI Live Product Audit.

The Inclusive AI Strategy Review

This review is designed to help organisations embed accessibility into their AI strategy, especially its governance and design processes. Our structured review addresses the risks associated with inaccessible AI, including legal exposure, bias, reputational damage, and the exclusion of disabled users.

Reviewing your AI practices against the EU AI Act, ISO 42001, ISO 30071-1, and WCAG 2.2 AA standards, ensure your AI initiatives and usage is compliant with all relevant AI and accessibility best practice.

Our review provides a comprehensive audit and a clear action plan, outlining how you can address any potential disability inclusion and accessibility gaps, and summarising for senior leadership teams how accessibility in AI is being championed in an ongoing way.

We’ll review:

1. Your AI base models and data sources: Ensuring your AI base model and data sources are comprehensive to mitigate bias and avoid gaps in understanding.
2. Decision making: Reviwing decision-making processes, as bias can lead to negative sentiment and unfair outcomes, particularly for edge cases involving disabled individuals.
3. Prompts: Implementing ‘guard rails’ to address unconscious bias in prompts, guiding the AI towards more inclusive responses.
4. Outputs: Carefully evaluating the outputs to ensure they are accessible and representative, while also leveraging AI for personalised user experiences.

The Inclusive AI Live Product Audit

Our Live AI Accessibility Audits are a dynamic service designed to assess the real-time accessibility of outputs and interfaces of AI-based products you’re creating or considering procuring. In a focused 2.5-hour session, we identify gaps in prompt design, output accessibility, and inclusive testing during AI development. This collaborative approach ensures that your AI systems are not only compliant but also user-friendly for all.