UN Secretary-General António Guterres welcomed the Call for Control of Frontier AI Models, launched by Finnish President Alexander Stubb and Norwegian Prime Minister Jonas Gahr Støre on 21 September, according to ANI’s report of his public statement. Guterres backed its appeal for UN member states to consider an international institution that could set standards, enable verification and bring governments together when AI systems cross capability thresholds.
Key points
- The call asks companies to establish safety protocols with pre-deployment testing and independent evaluation.
- It asks governments to coordinate standards and share reports of serious safety incidents.
- Guterres also encouraged frontier AI laboratories to engage with a UN scientific panel.
The Finland-Norway call puts testing before deployment
Stubb and Støre launched the initiative at United Nations Headquarters in New York on 21 September, WAM reported. The text calls for AI to remain under human direction, oversight and control, and for its development and use to comply with international law. It presents frontier models as offering potential benefits in science, economic activity and people’s lives while also posing safety and security risks if those risks are poorly managed.
For companies, the call seeks “transparent safety protocols, including mandatory pre-deployment testing and independent evaluation”. It also seeks sufficient access for qualified evaluators to assess risks. Those requests concern work before a model reaches users, as well as scrutiny by people outside the company developing it. They are proposals in an international appeal to companies, rather than obligations imposed by a regulator through this declaration.
Governments and regional organisations are asked to develop and coordinate common standards. The call also seeks greater transparency, including shared reporting of serious safety incidents, and access to scientific capacity, expertise and trusted evaluation for countries across all regions. Taken together, those requests would give governments a basis for comparing how developers assess risks, while extending evaluation capacity beyond the countries where frontier models are built.
UN members are asked to consider verification powers
The institutional proposal is directed at UN member states. It asks them to use existing international mechanisms as a foundation and consider an international institution to establish standards, support verification and bring states together once capability thresholds are reached. Guterres welcomed that approach in his public statement. A body with those functions would give states a place to agree on assessments and meet over a model’s capabilities, alongside the testing and evaluation sought from companies.
Leaders and senior officials from Australia, Bahrain, Canada, Denmark, Estonia, Germany, Iceland, Ireland, Kazakhstan, Kenya, Latvia, Moldova, the Netherlands, Singapore, Spain, South Africa, Türkiye and the United Arab Emirates endorsed the initiative, ANI reported. European Commission President Ursula von der Leyen also endorsed it.
The proposal gives verification a place alongside standards-setting. Its request that states be convened when capability thresholds are crossed would make a model’s assessed abilities a reason for governments to meet, rather than leaving discussion solely to a standing timetable. The call places that possible function within an institution member states are asked to explore. Its immediate requests to developers, meanwhile, concern protocols, testing and access for independent evaluators.
Guterres invites laboratories into the UN panel’s work
On 16 September, ahead of the 81st session of the UN General Assembly, Guterres described runaway artificial intelligence, the climate crisis and deepening inequality as “three existential threats”. He also spoke of AI’s potential in health, education and climate work, while warning that development was outpacing human understanding of its risks. He urged governments to protect their citizens and argued that international cooperation was indispensable.
Guterres also welcomed the release of the first thematic brief from the Independent International Scientific Panel on AI. The brief concerns AI safety and the risks of losing human control, WAM reported. He encouraged external experts, including people at frontier AI laboratories and AI safety institutes, to engage with the panel. That invitation puts organisations developing and assessing the models in contact with a scientific body intended to inform international discussion of their risks.
