Kodu Muud Historic Moment as the EU Parliament adopts the first ever piece of legislation on AI, however it falls short on human rights protections

Historic Moment as the EU Parliament adopts the first ever piece of legislation on AI, however it falls short on human rights protections

märts 18, 2024

Muud

On Wednesday 13^th March 2024, the European Parliament adopted a first of its kind piece of legislation on Artificial Intelligence (AI), namely the AI Act. Nevertheless, after two years of negotiations, the European Disability Forum is calling the regulation a “partial win on accessibility”.

The general consensus amongst EUD and the wider EU Disability Movement is that although this piece of legislation is a welcome step towards ensuring safe and trustworthy AI in the EU, the accountability and transparency framework for high-risk AI is too weak. Accordingly, it will not fully protect the rights of persons with disabilities, including deaf people.

In short, the Act places some very modest limits on the use of dangerous AI systems and, therefore, does not take a human-rights based approach. Furthermore, the text includes phrases such as “human-centric and trustworthy AI”, “in accordance with fundamental rights obligations”, and “non-discriminatory”. However, these are merely empty words without real corresponding actions outlined in the text. Instead, the needs of the AI industry are prioritised rather than human rights. Nevertheless, the text does leave opportunities for stronger protections at the national level.

EUD has conducted an analysis of the Act to determine the legal provisions in place to protect the rights of deaf people, as well as where it falls short of doing so.

Positive aspects:

· Requirements for assessing fundamental rights impact of high-risk AI systems (Article 27);

· Specific reference to the CRPD, highlighting the importance of universal design principles for the development of all AI systems in the Union (Recital 80);

· Obligations of Providers of high-risk AI systems to ensure they comply with accessibility requirements in line with EU Directives 2016/2102 and 2019/882 (Article 16);

· Member States are obliged to draw up codes of conduct concerning all AI systems such as assessing and preventing any negative impact on persons with disabilities and gender equality(Article 95);

· Prohibits AI systems that exploit any of the vulnerabilities of a person or a specific group of persons due to their disability (Article 5).

How the Act falls short:

· No provision ensuring data collection efforts for AI account for the diversity of disability e.g. deaf people (Article 10);

· Recognises persons with disabilities as a group at higher risk of harmful effects of AI systems, however there aren’t any concrete actions provided to combat these in the articles;

· Acknowledgement that the use of AI systems in education may violate the rights of persons with disabilities, including deaf people, however there are no concrete actions outlined in any article to prevent this (Recital 56);

· Recognition that AI systems used in employment, e.g. for recruitment purposes, should be classified as high risk and may discriminate against persons with disabilities, including deaf people, however there are no concrete actions outlined in any article to prevent this (Recital 57);

· Recognition that AI systems to evaluate credit score or worthiness are also high risk as they deal with access to finances of individuals which may well discriminate against persons with disabilities, including deaf people, however there are no concrete actions outlined in any article to prevent this (Recital 58);

· Acknowledgement of the importance of promoting the development of AI systems that improve accessibility of persons with disabilities, however there are no concrete examples of this any article (Recital 142);

· References to accessibility is phrased as follows – “…in a language which can be easily understood by the authority in one of the official languages of the institutions of the Union…” However, this excludes the EU’s national sign languages given it hasn’t granted them legal recognition on an equal footing with its spoken languages. Therefore, the Act neglects to include reference to accessibility of information and communication in the national sign languages (Article 21).

EUD will continue to advocate for the development and deployment of deaf-inclusive and non-discriminatory AI systems across the EU by working with its member organisations – National Associations of the Deaf – to carry out this action at the Member State level.