TLDR: Jade Leung, formerly the Governance Lead at OpenAI, has been appointed as the Chief Technology Officer of the UK AI Safety Institute. In this critical role, she is tasked with developing and overseeing evaluations for advanced AI models to ensure their safety and address potential risks, including those related to misuse in cyber, biological, or chemical attacks, and the challenge of AI systems maintaining human alignment and control.
The United Kingdom has significantly bolstered its artificial intelligence safety initiatives with the appointment of Jade Leung, a distinguished expert formerly with OpenAI, as the Chief Technology Officer (CTO) of the UK AI Safety Institute (AISI). Leung, who transitioned from her role as Governance Lead at OpenAI in October 2023, now spearheads the technical efforts at the AISI, a body rapidly establishing itself as a global leader in evaluating the safety of powerful AI systems.
At OpenAI, Leung was instrumental in developing secure AI systems, focusing on scaling and security protocols, particularly concerning artificial general intelligence (AGI). Her move to the public sector, despite reportedly involving a significant pay cut, underscores her commitment to shaping the governance of AI in the public interest. As she stated, “You really want a public interest body that is genuinely representing people to be making those decisions. By definition, they need to be [taken] outside of the organizations developing AI models.”
In her capacity as CTO, Leung’s primary responsibilities include building the nascent institute’s capabilities, staffing its technical teams, and, crucially, designing and overseeing rigorous evaluations of frontier AI models. These evaluations are designed to identify and mitigate dangerous capabilities and security vulnerabilities. Specific areas of focus include assessing the risk of AI systems being used to facilitate cyber, biological, or chemical attacks, as well as more speculative tests to determine if AI systems could potentially escape human control.
Under Leung’s leadership, the UK AISI has already made notable progress. The institute has successfully negotiated early access to cutting-edge models from leading AI companies for safety testing. Among its completed evaluations, the AISI has tested Anthropic’s Claude 3.5 Sonnet model. Leung asserts the institute’s leading position, stating, “We are currently, I would say, the best single place that you could go to to have your models evaluated across a range of frontier AI risk areas in a fairly robust and in depth way.”
Also Read:
- Advanced AI Models Achieve Expert-Level Proficiency in Privacy and Governance Certifications
- Igor Babuschkin, xAI Co-founder, Departs to Launch AI Safety-Focused Investment Firm
Leung’s extensive background includes co-founding the Centre for the Governance of AI at the University of Oxford, where she also completed her DPhil in International Relations as a Rhodes Scholar, focusing on AI governance. Her expertise and dedication to the field were recognized in 2024 when she was included in Time’s list of the “100 Most Influential People in AI.”
