Leading global artificial intelligence experts convened in London recently, including Turing Award laureate Andrew Chi-Chih Yao, who serves as dean of both Tsinghua’s Institute for Interdisciplinary Information Sciences and College of Artificial Intelligence; Turing Prize winner Yoshua Bengio; Zhang Ya-Qin, founding dean of Tsinghua's Institute for AI Industry Research; and UC Berkeley professor Stuart Russell, plus numerous other prominent AI researchers worldwide. These experts jointly signed the IDAIS London Declaration during their gathering. They also held discussions focused on stopping societal-scale harms stemming from AI-enabled attacks. The event marked the fifth session in the International Dialogues on AI Safety (IDAIS) series and was co-hosted by the Safe AI Forum and the University of Oxford's Martin AI Governance Initiative.
Front row from left: IDAIS founding members Stuart Russell, Andrew Chi-Chih Yao, Yoshua Bengio and Zhang Ya-Qin, pictured alongside fellow participating AI researchers in London.
Over the three-day conference, the assembled scientists and governance experts drafted and signed a consensus statement. It calls on states to recognize the common threat posed by the proliferation of AI-enabled cyberattacks and biological misuse capabilities and to commit to coordinated prevention efforts.
The statement warns that, on current trajectories, non-state actors with minimal technical capacity will gain access to certain state-level cyberattack capabilities within the next year, creating unprecedented risks to critical infrastructure and national security. Societies remain unprepared for this imminent threat, and biological misuse risks are not far behind.
The statement also sets out clear priorities across both risk areas, including protecting critical infrastructure, hardening safeguards on frontier models, and developing robust evaluation capacities. It cautions that what is now unfolding in the cyber domain should serve as a warning for biological risks and for the broader set of challenges ahead, including the potential loss of control over increasingly autonomous AI systems.
Top AI scientists gather for the fifth session in the International Dialogues on AI Safety (IDAIS).
List of signatories to the Declaration
The International Dialogues on AI Safety (IDAIS) brings together the world's leading scientists to jointly address various risks posed by artificial intelligence. The first dialogue, held in October 2023, was co-initiated by Andrew Chi-Chih Yao, Yoshua Bengio, Zhang Yaqin, and Stuart Russell. To date, IDAIS has held five international dialogues on AI safety: IDAIS-Oxford (2023), IDAIS-Beijing (2024), IDAIS-Venice (2024), IDAIS-Shanghai (2025), and IDAIS-London (2026).
Delegates from the Safe AI Forum and experts from Tsinghua University’s College of Artificial Intelligence hold a joint AI safety seminar.
In March 2026, Fynn Heide, CEO of the Safe AI Forum—the organizer of IDAIS—led a delegation to Tsinghua University’s College of Artificial Intelligence, where they engaged in academic exchanges and discussions with Dean Andrew Chi-Chih Yao and other experts on key AI safety issues.
Link to the IDAIS London Declaration:
https://idais.ai/dialogue/idais-london/
Editors: Li Han, John Paul Grima
