Tug of war continues on international AI treaty as text gets softened further
Last week, the negotiations on the world’s first treaty on Artificial Intelligence failed to solve the thorny issue of whether private companies should be covered. At the same time, the latest draft, seen by Euractiv, further watered down an already weakened text.
The Convention on Artificial Intelligence, Human Rights, Democracy, and the Rule of Law is a first-of-its-kind binding treaty developed in the context of the Council of Europe (CoE), a human rights body with 46 member states.
The United States and other countries like Canada, Israel, and Japan are not part of the CoE but participate in the discussion as observers, meaning they do not have voting rights but can influence the discussions by threatening not to sign the convention.
Since the start of the negotiations, Washington and these other countries, together with the United Kingdom, have been pushing to limit the treaty’s scope to public bodies by default, introducing the possibility for signatory countries to ‘opt-in’ their private companies.
Earlier this month, in a meeting of the Telecom Working Party, a technical body of the Council of EU Ministers, several countries asked the European Commission to show flexibility on the scope, considering the international uptake of the treaty should be given priority.
However, as Euractiv revealed, the Commission resisted pressure from the European governments and presented at the Committee on Artificial Intelligence’s plenary last week an ‘opt-out option’ meant to meet the reticence of the US administration.
The plenary was meant to reach a final decision on the scope issue, but it failed due to the lack of flexibility from both sides. Euractiv understands that only the EU’s opt-out and the US’ opt-in options are still on the table.
EU Commission’s last-minute attempt to keep private companies in world’s first AI treaty
Despite pressure from some EU countries, the European Commission is still trying to prevent private companies from being excluded by default from the first international treaty on Artificial Intelligence.
Consistent weakening
If the timeline of the ministerial adoption in May is to be respected, and it would have to be, as there might technically not be enough time to extend the Committee’s mandate if the negotiations fail, the next plenary in mid-March is set to be the decisive one.
However, the treaty’s provisions have been systematically weakened because the European Commission has pushed for a complete overlap with the AI Act, even when the human rights treaty could have gone beyond without conflicting with the EU product safety rules.
This alignment has meant that the EU is pushing to introduce in the convention carve-outs for national security, defence, and law enforcement that are even broader than in the AI Act, which might create loopholes for AI systems used for both civilian and military purposes.
If the treaty were to apply to only public bodies, these exemptions would make the scope extremely limited. At the same time, with each side trying to take out different bits and pieces, hardly anyone defends the original text.
EU prepares to push back on private sector carve-out from international AI treaty
The European Commission is preparing to push back on a US-led attempt to exempt the private sector from the world’s first international treaty on Artificial Intelligence while pushing for as much alignment as possible with the EU’s AI Act.
As a result, in a revised version of the text dated Friday (26 January) and seen by Euractiv, the weakening of the convention has reached a point that it is now closer to a declaration than a binding treaty.
For instance, in several key provisions like the ones to protect the democratic process or the procedural safeguards, the signatory party should “seek to ensure” that adequate measures are in place, a wording that implies no obligation.
Entire provisions, such as protecting health and the environment, measures promoting trust in AI systems, and the requirement to provide human oversight for AI-driven decisions affecting people’s human rights, have been scrapped.
Research activities have also been excluded from the treaty’s scope. Still, several options might extend this exception to the development phase, meaning no human rights protection would be introduced by design in the initial phase of the AI systems’ lifecycle.
In addition, concerning risk management, the disposition requiring the publication, even if only ‘where appropriate’, of the details of the risk analysis and mitigation measures has been deleted.
Political motives
According to several people involved in the AI convention, who requested anonymity due to the confidential nature of the discussions, political considerations were a key explanation for why observer countries like the United States were allowed to play such a determining role in shaping the treaty.
The political motives are made more evident by the fact that, while the US administration might sign the treaty, there is close to zero possibility that Congress would ever ratify it, which is needed to make it legally binding.
EU’s AI ambitions at risk as US pushes to water down international treaty
Just as the timing of the world’s first AI treaty starts aligning with the EU legislative agenda, an American-led push to exclude private companies might make it not worth the paper it is written on.
The Council of Europe is an …
Still, Washington could claim that they are taking steps to address AI-related human rights violations while preventing setting a dangerous precedent with binding language that might be put on the table in other fora like the United Nations.
At the same time, the American signature might be seen as a major diplomatic victory for the Council of Europe, which would be the home of the first genuinely international treaty on Artificial Intelligence at a time when the race for the institution’s top job is up for grabs.
The diplomatic victory would also go to Switzerland, as the chair of the AI Committee is a Swiss, Thomas Schneider. Coincidentally, one of the key candidates for the CoE’s top job is former Swiss President Alain Berset.
The sources familiar with the discussions confirmed that the chair and the CoE’s secretariat were not neutral throughout the negotiations, instead pushing for the arguments of the US and other observer countries while putting aside contradicting ones.
“An international organization whose purpose is to ensure human rights protection but which issues a treaty without any commitment on the private sector, from where most violations originate, just loses any credibility and jeopardises its legitimacy,” one of the people involved told Euractiv.
Read more with Euractiv
