Recently, Yao Qizhi, a Turing Award laureate and Dean of the School of Artificial Intelligence at Tsinghua University, joined forces with fellow Turing Award winner Yoshua Bengio, Zhang Yaqin, Dean of the Institute for AI Industry Research at Tsinghua University, Stuart Russell, a distinguished professor at the University of California, Berkeley, and other leading AI experts from across the globe, to co-sign the 'IDAIS London Declaration' at the Royal Society in the United Kingdom. This event marked the fifth installment of the International Dialogue on AI Safety (IDAIS) series, co-organized by the Forum on Secure Artificial Intelligence and the Martin School of Governance for Artificial Intelligence at the University of Oxford. The primary objective of this gathering was to proactively avert widespread societal harm stemming from AI-driven attacks.

The declaration highlights the looming threats posed by AI-enabled cyber and biological abuses. It underscores the alarming possibility that non-state actors could acquire cyberattack capabilities on par with national entities within the next year, and notes that the risk of AI-facilitated biological abuses is also on a steady ascent. In response to these imminent dangers, the declaration advocates for concrete measures, including bolstering the protection of critical infrastructure, establishing robust evaluation mechanisms for AI attack capabilities, and implementing rigorous pre-release testing of AI models.