The French Summit pioneered a format similar to climate change's COPs, featuring not only the main summit but also hundreds of side events focused on AI topics. The wide array of events organized alongside the main summit were a major highlight of the Summit. Here’s a wrap up regarding SaferAI’s interventions.
Our Research Lead Malcolm Murray spoke about risk thresholds at a roundtable organized on the Science Days, prior to the main Summit. He spoke alongside other experts like Frontier Model Forum’s CEO Chris Meserole, an EU AI Office representative and a nuclear energy expert. At this event, he emphasized the importance of setting risk thresholds and striving to ground them in concrete risk estimates, rather than relying only on indirect proxies. This is a core motivation of our recent work attempting to turn empirical measurements into risk estimates (Murray et al. 2025).
On the same day as the Science Day roundtable, our colleague James Gealy was invited to intervene at the OECD during the launch of the G7 Hiroshima AI Process in which SaferAI was the only NGO to have participated. Speaking alongside representatives from all G7 countries, he highlighted the commonalities between the G7 Hiroshima Process and the EU Code of Practice, particularly the shared focus on comparable systemic risks.
We successfully supported the organization of a high profile dinner with Tony Blair Institute & Faculty AI. In their opening statement, both Peter Kyle (UK Minister of Science and Technology) and Josephine Teo (Singapore’s Minister for Digital Development and Information) expressed a stance of cautious optimism about AI, highlighting the potential of the technology and the will of their country to lead in it, while acknowledging the tremendous risks arising from those. The dinner also revealed a notable consensus on the need for global coordination on risks, voiced by both AI pioneer Yoshua Bengio and former Chinese Vice Minister of Foreign Affairs, Madam Fu Ying.
Our CEO attended the core Summit on the Monday. This allowed him to meet with several leading figures in the field. Most notably, it also enabled a brief conversation between our CEO and President Macron. After expressing support to the president on the development of a “non-aligned” strategy with respect to AI, our CEO remarked that the risks of extinction were very real, as stated by multiple AI Nobel laureates. After a brief exchange, President Macron indicated that the question remained up for debate, to which our CEO followed up, without success to date.
Our Policy Lead, Chloé Touzet, represented civil society at this event, speaking alongside experts from the Frontier Model Forum, Google DeepMind, METR, and AWS. She emphasized the importance of transparency for governments to be aware of the frontier and be able to monitor risks as those increase.
Henry Papadatos, our Managing Director, spoke about the involvement of civil society in AI risk management at a roundtable alongside Renaissance Numérique, The Future Society, Wikimedia France and the Avaaz Foundation.
On the 11th of February, we co-organized a series of roundtables with OECD, the UN Human Rights Office, and the AI Standards Hub, on AI risk management, at the OECD. As a part of this series, our Policy Lead ran a roundtable which was the occasion to compare core elements from risk management frameworks outlined by the EU Code of Practice, the Frontier Model Forum and civil society organizations like ourselves. It is encouraging to see that a consensus is starting to form around the core building blocks that AI risk management frameworks should have, i.e. setting risk thresholds, risk identification, risk analysis & evaluation, risk mitigation and risk governance.
