The Convention on Artificial Intelligence, Human Rights, Democracy, and the Rule of Law, developed by the human rights body, the Council of Europe, is a binding treaty involving 46 member states of the EU, along with the US Canada, Israel, and Japan, acting as observer countries.

While the Council of Europe aims to extend the treaty scope to cover private entities, the US and other observer countries are strongly advocating for the treaty's scope to be limited to public bodies only, with the option to 'opt-in' for private companies for the signatory countries. This has resulted in a stalemate in negotiations, highlighting the lack of consensus among the nations. Many have raised concerns over the influence of the observer countries in the treaty process, with the European Commission putting forward an 'opt-out option' for private countries to address the concerns of the US administration. However, the plenary meeting has failed to reach a decision in this matter. Critics have argued, that without the inclusion of the private sector, where most AI research and development activity is still undertaken, the AI treaty loses its credibility and legitimacy.

Meanwhile, the treaty text is consistently getting diluted, raising questions over its overall efficacy and making the treaty closer to a declaration than a binding agreement. The European Commission is seeking to align the treaty with the EU AI Act and this has led to carve-outs for national security, defence, and law enforcement in the treaty, that are much broader in scope than those in the AI Act. This alignment has produced legitimate fears that it may provide impetus for the creation of loopholes in AI systems, applicable in both civilian and military contexts. Furthermore, research activities in AI development are excluded from the purview of the treaty and even critical provisions, such as those involving the protection of health and the environment, promotion of trust in AI systems, and human oversight requirements, have been removed.

Read more: Euractiv