Stay updated with the latest news through complimentary updates by subscribing to myFT Digest, an AI newsletter that will be conveniently delivered to your inbox.
In a groundbreaking move, the US and the UK have joined forces to address the analysis and evaluation of threats posed by emerging AI models. The collaborative agreement, signed in Washington, DC, brings together the technical expertise and skills of both countries in the realm of AI safety. Michelle Donelan, the British technology minister, and Gina Raimondo, the US commerce secretary, formalized this historic pact on Monday.
This bilateral agreement marks a significant milestone in the history of AI safety, as both nations aim to enhance oversight of potential existential threats arising from advanced technologies. The accord paves the way for cooperation on critical issues such as cybersecurity vulnerabilities and the misuse of AI in developing bioweapons.
Emphasizing the urgency of the situation, Donelan highlighted the transformative potential of the next generation of AI models, stressing the need for swift action to understand and harness their capabilities effectively.
Under this agreement, the UK’s recently established AI Safety Institute (AISI) and its US counterpart will engage in knowledge exchange through the exchange of analysts. This collaboration may extend to the evaluation of proprietary AI models developed by leading companies like OpenAI and Google.
Drawing inspiration from the close collaboration between the UK’s Government Communications Headquarters (GCHQ) and the US National Security Agency, this partnership sets a precedent for shared intelligence and security initiatives.
Donelan underscored the importance of the UK’s expertise in AI safety, positioning the country as a trailblazer in shaping regulatory frameworks and fostering innovation in the AI sector. While advocating for rigorous studies on AI protection, she acknowledged the challenges of regulating a rapidly evolving technology landscape.
In contrast to global trends, the EU has enacted the stringent AI Act, while US President Joe Biden has issued executive orders targeting AI principles with potential implications for national security. Meanwhile, China has introduced guidelines to safeguard its regulatory regime against AI disruptions.
Raimondo described AI as the cornerstone of modern systems, emphasizing the collaborative efforts between the US and the UK in addressing a spectrum of risks spanning national security and societal impacts.
This strategic partnership aims to deepen the understanding of AI systems, conduct robust evaluations, and provide comprehensive guidance through the engagement of researchers from esteemed institutions like the University of Oxford and industry leaders such as Google DeepMind.
Notable tech companies including OpenAI, Google DeepMind, Microsoft, and Meta have voluntarily committed to subjecting their latest AI designs to evaluation by the UK’s AISI. This initiative aligns with Prime Minister Rishi Sunak’s vision for the UK to play a pivotal role in shaping the future of AI development.
By leveraging the expertise of the National Cyber Security Centre at GCHQ, the testing of AI designs addresses critical cybersecurity concerns associated with technological advancements.
Donelan and Raimondo are set to address shared concerns, including the impact of AI on upcoming elections, and explore the potential for AI-driven computing infrastructure to benefit the public through enhanced collaboration and knowledge-sharing initiatives.
