The UK AI safety institute aiming to strengthen global AI safety partnerships and capitalize on Bay Area technological talent, expands to the United States.
The Artificial Intelligence (AI) Safety Institute of the United Kingdom is preparing to establish a new branch in the United States as part of its international expansion.
On May 20, Michelle Donelan, the UK Technology Secretary, announced that the institute will establish its first overseas office in San Francisco in the summer.
The announcement said that the strategic choice of a San Francisco office would enable the U.K. to “tap into the wealth of tech talent available in the Bay Area,” along with engaging with one of the world’s largest AI labs located between London and San Francisco.
Additionally, it said this move will help it “cement” relationships with key players in the U.S. to advocate for global AI safety “for the public interest.”
The London branch of the AI Safety Institute already has a team of 30 on a trajectory to scale and acquire more expertise, particularly in risk assessment for frontier AI models.
Donelan said the expansion represents the U.K.’s leader and vision for AI safety.
“It is a pivotal moment in the UK’s ability to study both the risks and potential of AI from a global lens, strengthening our partnership with the US and paving the way for other countries to tap into our expertise as we continue to lead the world on AI safety.”
This follows the UK’s landmark AI Safety Summit in London in November 2023. The summit was the first in concentrating on AI safety on a global scale.
The event boasted of leaders from around the globe, including the U.S. and China, and leading voices in the AI space, including Microsoft president Brad Smith, OpenAI CEO Sam Altman, Google and DeepMind CEO Demis Hassabis and Elon Musk.
In this latest announcement, the UK also disclosed a selection of the institute’s recent results from safety testing conducted on five publicly available advanced AI models.
It anonymized the models and said the results provide a “snapshot” of the capabilities of the models instead of designating them as “safe” or “unsafe.”
Some findings showed that several models could complete cyber security challenges, while others labored with more advanced ones. Several models were found to have PhD-level knowledge of chemistry and biology.
It concluded that all tested models were “highly vulnerable” to basic jailbreaks and that the tested models could not complete more “complex, time-consuming tasks” without human supervision.
Ian Hogarth, the institute’s chair, said these assessments would help to appraise model capabilities empirically.
“AI safety is still a very young and emerging field. These results represent only a small portion of the evaluation approach AISI is developing.”
