ADAPT Centre Contribution on Implementation of the EU AI Act and Fundamental Right Protection

The EU AI Act introduces a blanket protection of fundamental rights for specific applications of AI that it classifies as high-risk, which is implemented under the existing single market harmonised product certification mechanisms for health and safety protection, i.e. the New Legislative Framework. This protection of fundamental rights places many AI issues previously covered by voluntary trustworthy or ethical AI frameworks into a framework with independent and legally binding accountability for harmful characteristics of products grounded in the same human rights framework underpinning Union Law and many national laws. However, this major change in accountability also introduces many legal uncertainties on how AI providers and deployers can identify and manage risks to fundamental rights. Contrast this to the introduction of GDPR, which focussed on the protection of rights of privacy and data protection but benefitted from the development and employment of data protection principles under the data protection directive which had been in force beforehand. The protection of fundamental rights in AI systems however benefits from no such breakdown of principle, nor from prior deployment or compliance experience with such principles. This presents an extremely high level of legal uncertainty for providers and deployers of AI systems once the Act comes into force. The associated burden or chilling effects may fall disproportionately on public bodies wishing to deploy and reap the benefits of AI in high risk areas, and indigenous companies and especially SMEs that wish to market products into such applications.