European Parliament’s watering down of protections for children in AI Act risks worsening systemic exploitation of children online

The European Parliament’s leading committees on the AI Act adopted their final position on the proposed law. Despite the Parliament’s stated strong commitment to implementing children’s established rights in the digital environment, as set out in particular in its own report of the Special Committee on Artificial Intelligence (AIDA), the resultant legislative text falls short of ensuring that AI systems in the EU are safe for children, by design and default. In its position, the Parliament has not only failed to enhance protections of children and their rights: MEPs have effectively weakened them to the extent of making any duty of AI developers and operators to consider children’s rights and vulnerabilities practically inexistent.

While the AI Act claims to ensure a human-centric and ethical development of AI in Europe, based on EU values and reasons of public interest such as health, safety and fundamental rights, and the AIDA report stresses that “the implications AI has for children’s privacy, safety and security fall across a wide spectrum” and “that special consideration and protection need to be given to children when developing AI systems because of their particularly sensitive nature and specific vulnerabilities”, thus committing to children’s rights as set out in the UN Convention on the Rights of the Child and its General Comment 25, the latest position of the Parliament ditches a ban proposed by the European Commission on AI systems that exploit the vulnerabilities of children, replacing it instead with a more general one and setting a threshold for harm so high as to be meaningless.

The AI Act then sets out a list of AI systems that are considered high-risk and will be subject to strict due diligence measures, but MEPs have failed to retain AI systems likely – or even intended – to affect children’s safety or personal development on the high-risk list.

Without specific, enforceable provisions to ensure AI systems and AI regulators recognise children and take into account their needs and rights, the AI Act will not just fail to achieve its own aims. It will legitimise the ongoing and increasing systemic exposure of children to risks and harm online. The European Parliament must urgently review its position in view of negotiations with the Council of the EU on the law, to ensure the future of the digital world is one that serves those it will be left to.