Prime Minister Rishi Sunak’s ambition to establish a new international scientific body focusing on the risks associated with powerful AI systems has encountered a setback, according to a final summit communiqué obtained by POLITICO. While acknowledging the potential for “serious, even catastrophic, harm” from advanced AI, participating nations are leaning toward further scientific study through existing collaborative efforts, such as the United Nations (U.N.) and the Global Partnership on AI.
In a document dated October 25, presented by the General Secretariat of the European Council to EU member states, a notable change in wording is observed. The final version of the communiqué suggests the establishment of a network that “encompasses and complements” existing AI safety efforts, notably mentioning the U.N. This adjustment marks a departure from an earlier draft, which merely indicated that the network should “support” multilateral collaboration without specifying the U.N.
This shift in language may be seen as a setback for the U.K., which had aspired to create a novel global AI research body during its AI Safety Summit on November 1 and 2.
The U.N. had previously announced the formation of a high-level advisory body on AI in August, with expectations to name appointees soon. This body is tasked with presenting a final report by September 2024. The increased reference to the U.N. in the final communiqué suggests a preference for entrusting AI safety discussions to established international organizations.
U.K.’s response and ambitions
A U.K. official closely involved with the summit has challenged the interpretation of the new wording, asserting that it does not signify a scaling back of ambitions. The official, who spoke on condition of anonymity, emphasized that their intention, based on received feedback, was quite the opposite. However, this perspective contrasts with the altered language in the communiqué.
In another twist, an earlier reference to “future international AI Safety Summits” now appears within square brackets, indicating a lack of consensus among governments on this matter. The U.K.’s technology secretary expressed hope that the Bletchley Park summit would be the “first of many.”
International relations and omissions
Notably, references to UNESCO’s Recommendation on the Ethics of AI and the G20 have been removed from the final communiqué. While the reasons for these omissions remain unclear, it is worth noting that China is a member of both organizations. This raises questions about the geopolitical dynamics influencing discussions on AI safety.
The final communiqué places emphasis on proportionate governance policies. It suggests that countries may consider making risk classifications and categorizations based on their national circumstances and applicable legal frameworks. Additionally, the document highlights the importance of cooperation on approaches such as common principles and codes of conduct.
Alignment with European Union’s approach
Another document from the General Secretariat of the European Council indicates that feedback from an EU working group sought to align the statement more closely with the approach to AI pursued by the European Union. This alignment appears to be influenced by the ongoing work on the AI Act within the EU.
The international agreement on establishing a new AI safety body, championed by Prime Minister Rishi Sunak, has encountered obstacles. The altered language in the final summit communiqué, highlighting the role of existing organizations like the U.N., suggests a preference for leveraging established international efforts in AI safety research. While the U.K. government maintains that this change does not signal a scaling back of ambitions, the altered wording and omission of references to key international bodies raise questions about the evolving landscape of AI governance and international cooperation. The geopolitical implications of these developments remain a subject of scrutiny as the AI safety landscape continues to evolve.
