Every day we read more and more about AI and the impact on everyone’s life and privacy. Much of this gives us food for thought, particularly when it comes to protecting AAC users.  Policy has lagged behind developments, and much needs to be said about techno-ethics, AI and AAC.

The European Union (EU) Artificial Intelligence (AI) Act (2026)

In August 2026, the European Union (EU) introduced the Artificial Intelligence (AI) Act.  It covers the development, supply, and use of AI technology for citizens.

While the legislation is only relevant to the EU, it may affect AAC users worldwide.  This is because it covers devices developed and supplied from the EU wherever they are used.  It also covers devices made elsewhere and used in the EU.

Section 92 of the EU AI Act (2026) states: ‘Affected persons’ vulnerability: Certain groups, such as children, the elderly, or persons with disabilities, may be more susceptible to harm from specific AI systems. What may be considered less significant harm for persons in general might be considered significant and unacceptable for such vulnerable groups, especially children.’

What might this legislation mean for AAC users worldwide?

There is much to commend the developers and suppliers of AAC devices for; however, as AI continues to evolve, developers and suppliers must safeguard ‘vulnerable’ AAC users by ensuring they are not unfairly disadvantaged or put at risk through the use of AI. This might include:

  • How AI in large language models (LLMs) through machine learning (ML) supports word, phrase and sentence prediction and abbreviation expansion. Examples of disadvantages might include:
    • Moving from a dedicated LLM used by AAC users to adopting a mainstream LLM that does not support what disabled people wish to say without more effort to spell out individual words that are not predicted.
    • Developers and suppliers not checking that the LLM used by AAC users remains of high quality and disability friendly. This means monitoring for language evolving away from what is needed.
  • Using an AI function that predicts responses for an AAC user may put more vulnerable AAC users at risk by providing responses they would not have chosen themselves, raising authenticity-of-authorship questions. The legislation specifically appears to require AI-generated content to be flagged and identifiable.
  • Allowing a simple opt-in/opt-out AI button or feature that might be overridden inadvertently or misunderstood by an AAC user; a simple tick for ‘I am over xx years’ is not enough.
  • Tracking when an AI algorithm changes a data set, ensuring there is an audit trail to audit and explain the change(s) to prevent disadvantaging AAC users.
  • Providing an AI voice for an AAC device that then attracts a subscription or can only be used online.  What can be more harmful than being silenced?

How this legislation works with other existing legislation

There is also an interplay with other legislation. For example, Section 108 states: ‘AI applications that are not designed in an accessible manner should not be regarded to exploit vulnerabilities of persons with disabilities since they do not specifically target those vulnerabilities but are simply inaccessible to the persons with disabilities.’

Co-creation, co-production or consultation?

Developers drawing on mainstream LLMs and relying on AI machine learning already risk exploiting vulnerabilities or designing resources that leave behind, or at the very least sideline, AAC users, and this risk will only increase as AI continues to evolve at pace. For me, this once again brings into focus the crucial need for developers to work with AAC users. This cannot be in just a ‘consultative after the event’ manner, this has to be a right from the start first step mindset.  What do AAC users need and want?  In every walk of life, unless you have lived experience, it is challenging to understand the needs of those with disabilities.

Accessibility is, of course, a whole other debate for AAC, but combined with this AI legislation, it suggests there is now an urgent need for AAC developers and suppliers, alongside AAC users and policy professionals, to jointly create an industry-wide global policy and code of ethics that puts and keeps AAC users at the centre of device, software and AI developments. The question is: who should lead this, and who should monitor the effectiveness of the policy and techno-ethics, AI and AAC?