The world’s first international agreement on artificial intelligence is at risk of dilution. The US is now demanding that private companies be excluded from the deal, turning it into a toothless tiger.
From the beginning of 2022, all 46 member states and observer states of the Council of Europe are involved in the Artificial Intelligence Group. It deals with the drafting of a treaty on artificial intelligence, human rights, democracy and the rule of law.
For the EU, this binding international agreement could be a benchmark for its own regulation on artificial intelligence, the so-called AI Act.
However, Europe’s ambition to be the world’s digital leader collided with a practical problem: the Council of Europe’s work on the AI conference began before the EU. Negotiations in the Council of Europe therefore became difficult.
Last October, EURACTIV revealed that the European Commission, which is competent to negotiate on behalf of the EU, had managed to buy some time. The timetable has been delayed again and is now fully aligned with EU legislation on artificial intelligence, which is due to be finalized by the end of the year.
The September plenary session of the AI Committee, which was supposed to adopt the final text of the AI Treaty, has been cancelled. Adoption at committee level is now expected in March or April next year, followed by a ministerial decision in May or June.
But now that the question of the timetable appears to have been resolved, another glaring question stands in the way of the EU Commission’s ambitions, meaning the entire deal could water down under pressure from Washington.
Little commitment to transparency
In January, EURACTIV revealed that the United States had requested and that the drafting of the text took place behind closed doors, with only those who could sign it, i.e. the country’s representatives, able to attend. This often excluded interest groups such as civil society groups and private actors from the process.
The reason for this secrecy is this internal policy According to the Council of Europe, the participating countries do not want to make their positions public, especially since the United States has been pushing from day one to limit the scope of the agreement.
Backed by the UK, Canada and Israel, the US government has pushed to limit the scope of the AI conference to public companies and exclude the private sector. Rather, it implies Order “A cross-cutting legal instrument” of the group.
Despite EU opposition, the possibility of excluding private companies remains on the table. The US government refined its proposal during the committee’s plenary session in Strasbourg last week and presented its option for voluntary expansion.
“Each Party may, at the time of signature, when depositing its instrument of ratification or accession, or at a later date, declare the extent to which this Convention falls within its competence in relation to the design, development and deployment of artificial intelligence systems. It shall also apply on behalf of the Party,” states the US proposal seen by EURACTIV.
In other words, the Convention applies only to public bodies. Nevertheless, individual countries may decide whether an international agreement should apply to private companies within their respective national jurisdictions and declare this publicly.
This approach is a bit unusual because usually, if the treaty allows, countries declare they have reservations or withdraw from certain areas. However, a voluntary expansion option would avoid considerable public embarrassment for the government.
Although the US proposal would mean weakening the AI convention below Council of Europe standards, the group’s secretariat has so far jumped on the idea to ensure possible participation in the agreement, according to EURACTIV.
According to an informed source who spoke to EURACTIV on condition of anonymity, one reason for this soft approach could be the European Council’s hope that Washington will fund initiatives related to the AI treaty. This is currently the case with the Budapest Convention on Cybercrime.
Additional open questions
However, other aspects could lead to a weakening of the first international agreement on artificial intelligence. Although national security is already excluded from the current draft, participating countries are now discussing a broader exclusion of national security.
Another important point is that some governments are pushing for Council of Europe-approved data protection and non-discrimination to become common “principles”, which would leave countries more room for interpretation.
Ultimately, how robust the monitoring mechanism should be and whether external experts should be involved is also controversial. A key part of this problem is that some countries prefer discretion in restricting access to remedies.
[Bearbeitet von Alice Taylor/Kjeld Neubert]
“Amateur coffee fan. Travel guru. Subtly charming zombie maven. Incurable reader. Web fanatic.”