AI, Disability and Inclusion: Is Artificial Intelligence Opening Doors or Creating New Barriers?
The AI Funding Files: Following the Money, Examining the Impact
Artificial intelligence is increasingly being used to support communication, learning, creativity and everyday tasks. For disabled and neurodivergent people, these developments could create new opportunities for independence, self-expression and participation.
AI can describe images, convert complicated information into simpler language, assist with writing, generate captions and help people explore music, art and storytelling. But there is an important question behind the excitement: are these technologies being designed around the people who need them, or are disabled people being expected to adapt to systems that were never built with them in mind?
The answer matters because technology alone does not create inclusion. Accessibility, affordability, human support, choice and respect for individual rights matter just as much.
1. Where AI could open doors
When designed and used responsibly, AI can help remove some of the practical barriers people encounter in education, work, communication and creative activities.
- Communication: speech-to-text, text-to-speech, captioning and tools that help people organise or express their thoughts.
- Accessible information: converting complicated material into clearer language, producing summaries and supporting alternative formats.
- Visual access: generating descriptions of images and helping people navigate visual information.
- Learning: offering different explanations, adjustable practice activities and support at an individual's preferred pace.
- Creative participation: helping people develop song lyrics, make music, explore artwork, write stories, edit videos and experiment with new ideas.
- Everyday independence: supporting planning, reminders, organisation and the completion of tasks, where the tool is appropriate and reliable.
These possibilities are especially relevant to inclusive creative projects. Someone who finds writing difficult might express an idea verbally and use an AI tool to organise it. A person who communicates differently might use assistive technology to contribute to a group project. An emerging musician might experiment with production tools without needing access to an expensive studio at the beginning.
However, AI should support a person's contribution rather than erase it. The individual's preferences, ideas, decisions and creative ownership should remain central.
2. The danger of creating new barriers
Technology can reproduce exclusion when developers do not consider the full range of human needs.
A speech-recognition system may struggle with a person's speech patterns, accent or communication device. An AI-generated explanation may misunderstand a person's support needs. A recruitment system may disadvantage someone whose communication or employment history differs from the examples used to develop it.
Other barriers include:
- Interfaces that do not work well with screen readers, keyboards or alternative input devices.
- Tools that assume everyone processes information, communicates or learns in the same way.
- Subscription costs, inaccessible equipment or unreliable internet access.
- Systems that make assumptions about someone's abilities based on incomplete or biased data.
- Services that replace personal support with automated responses without checking whether this meets the individual's needs.
- Limited opportunities for disabled people to report problems, challenge decisions or influence product development.
The UK's Information Commissioner's Office warns that AI can produce discriminatory effects when its training data, design or use reflects existing inequalities. The Equality Act 2010 also remains relevant where organisations provide services, employ people or make decisions affecting disabled people.
It is therefore not enough for a company to say that its technology treats everybody the same. A system may apply the same process to everyone and still place some people at a substantial disadvantage.
3. Accessibility must be designed in, not added later
Accessibility should be considered from the earliest stages of developing or buying an AI product. It should not be treated as an optional improvement to be considered after the main product has been completed.
Good practice includes involving disabled and neurodivergent people in research, design, testing and evaluation. Organisations should test tools with people who use assistive technology and people with different communication, sensory, cognitive and physical access needs.
They should also provide alternatives when an AI system is unsuitable. People should not be forced to use a particular automated tool as the only route to education, employment, public services or support.
For public bodies in England, the Equality and Human Rights Commission's guidance on AI in public services explains the importance of considering equality duties when using these systems. Accessibility obligations also continue to apply to covered public-sector websites, documents and services, including where AI-enabled technology is involved.
4. Following the money: what should funding support?
AI investment is often discussed in terms of innovation, productivity, new products and economic growth. Disability inclusion requires a wider view of what counts as a successful investment.
Funding an AI product is not the same as funding accessible outcomes. A project may develop impressive technology but fail to reach the people who could benefit from it because there is no money for accessible design, training, equipment, ongoing support or independent evaluation.
Investment decisions should consider the complete journey from development to real-world use.
Five areas that deserve attention
- Accessible design and testing: funding disabled people's involvement in development, usability testing and improvements.
- Access to equipment and connectivity: helping address the costs that prevent people from using digital services.
- Training and human support: enabling disabled people, families, educators, employers and support workers to use appropriate tools safely.
- Independent evaluation: assessing whether a product improves access in practice, including for people whose needs are often overlooked.
- Long-term maintenance: supporting updates, technical assistance, privacy protection and continued access after a pilot or grant ends.
The UK Government's Digital Inclusion Action Plan, published in February 2025, sets out work to address the digital divide. Its existence is an important policy starting point, but a policy announcement should not be mistaken for proof that every disabled person has gained access or that every proposed activity has received funding.
Internationally, the United Nations' 2025 Human Development Report examines AI's potential to support people with disabilities while warning that inaccessible or unsuitable technology can deepen marginalisation. Its wider message is significant: technological capability does not automatically translate into equal participation.
For anyone assessing an AI funding announcement, the questions should be practical: How much money has actually been allocated? Who can apply? Are disabled-led organisations eligible? Is there funding for access needs and support? What results will be measured, and will the findings be published?
5. Education, employment and independence
AI tools may help some learners access material in different formats, practise skills or communicate their understanding. In employment, they may assist with drafting, organisation, transcription and other tasks.
But these benefits depend on how the tools are implemented. An education provider should not assume that an AI-generated explanation is appropriate for every learner. An employer should not rely unquestioningly on automated recruitment scores. A support service should not allow a chatbot to become a substitute for human judgement when a person needs individual attention.
Where AI influences decisions about an individual, organisations need to consider fairness, transparency, privacy and the ability to challenge errors. Human oversight must be meaningful: a person should have the knowledge, authority and time to question a system rather than simply approve whatever it produces.
6. AI and creative inclusion: who gets to make culture?
Creative technology offers a particularly important opportunity. Disabled and neurodivergent people should be able to participate in music, film, writing, visual art and digital storytelling as creators, not merely as the subjects of other people's work.
AI can be one tool in that process. It might help someone explore a melody, develop a beat, generate visual ideas, organise a script or experiment with different production techniques. It can also provide new ways to collaborate when traditional creative environments create barriers.
Yet creative inclusion requires more than providing access to software. Participants should be able to choose how they take part, decide what they want to share, receive appropriate credit and understand how their work, image, voice and personal information may be used.
Consent should be informed and specific. A person agreeing to use AI during a workshop has not automatically agreed to have their voice cloned, their image used in promotional material or their work included in a public release. Where someone needs communication or decision-making support, that support should enable their own choices rather than replace them.
For inclusive music and media projects, a strong approach combines accessible creative tools with safeguarding, clear agreements, human support and respect for artistic ownership.
7. Privacy, dignity and safeguarding
AI tools may process sensitive information, including personal stories, images, recordings, communication samples and information about someone's disability or support needs. Organisations should understand what information a tool collects, how it is stored, whether it may be used to improve the service and who can access it.
Before introducing AI into a disability service, school, workplace or creative project, organisations should establish clear rules for consent, data minimisation, retention, access and incident reporting. They should avoid entering identifiable or confidential information into tools that have not been approved for that purpose.
Safeguarding must also extend to synthetic media. AI-generated images, voices and videos can be used in ways that misrepresent a person or undermine their dignity. Projects should have clear permissions for recording and publishing, and a process for raising concerns or requesting that material be reviewed.
People should not have to surrender privacy or control over their identity to benefit from accessible technology.
8. A practical checklist for organisations
Before buying, funding or introducing an AI tool, organisations can ask:
- Have disabled and neurodivergent people helped choose and test the tool?
- Does it work with relevant assistive technology and alternative communication methods?
- Can people use another route if the AI system is unsuitable?
- Are costs, training and ongoing support included in the budget?
- What personal information is collected, and is the use lawful, transparent and proportionate?
- Could the tool reproduce bias or unfairly disadvantage particular groups?
- Who is accountable when it gives incorrect advice or makes an error?
- Can a person request a human review or challenge an important decision?
- Are consent, creative ownership and the right to withdraw from a project clearly explained?
- Will the organisation publish evidence of outcomes, including problems and lessons learned?
9. What should meaningful progress look like?
Success should not be measured only by how many AI tools an organisation purchases or how many people attend a technology demonstration.
More meaningful measures include whether people can participate more independently, communicate more effectively, access opportunities previously closed to them, exercise greater control over their work and challenge decisions that affect them.
Evaluation should include the experiences of disabled people themselves. Organisations should report who benefited, who did not, what barriers remained and what they changed as a result. Where a project uses public money, transparent reporting can help communities understand whether the investment delivered its stated aims.
Conclusion: inclusion is a responsibility, not a software feature
Artificial intelligence could help remove barriers in communication, education, employment and creative expression. It could also reinforce existing inequalities if accessibility, fairness and human rights are treated as afterthoughts.
The distinction lies in the choices made by developers, funders, employers, educators, public services and creative organisations.
For AM-MediaSpace, responsible AI means treating technology as an assistant rather than an unquestioned authority. It means protecting people's rights, checking important information, maintaining meaningful human oversight and creating space for disabled and neurodivergent people to shape their own stories.
The question is not simply whether AI can do more. It is whether more people can participate, create and make decisions on their own terms.
The funding test: Does the investment give disabled people more choice, access and control — and can the organisation show the evidence?
Sources and further reading
- UK Government — Digital Inclusion Action Plan: First Steps (2025)
- UK Government / Equality and Human Rights Commission — Artificial Intelligence in Public Services
- Information Commissioner's Office — AI Fairness, Bias and Discrimination
- UK Government — Accessibility of Private-Sector Products and Services for Disabled People in the UK (2025)
- UK Government — Data and AI Ethics Framework
- United Nations Development Programme — Human Development Report 2025
- Equality and Human Rights Commission — Disability Discrimination
Editorial note: This article discusses policy, accessibility and responsible technology. It does not claim that every policy commitment has been fully funded or implemented, and it is not legal advice.
Comments