The UK and US have signed a groundbreaking agreement to cooperate on developing rigorous testing of advanced AI systems, a major step forward in ensuring their safe deployment.
The memorandum of understanding signed Monday by UK Technology Secretary Michelle Donnellan and US Commerce Secretary Gina Raimondo establishes a partnership to align the two countries' scientific approaches in rapidly iterating robust evaluation methodologies for cutting-edge AI models, systems and agents.
Under the agreement, the new UK AI Security Institute and the upcoming US organisation will exchange research expertise, including how to independently assess civilian AI models from companies like OpenAI, with a view to mitigating AI risks. The partnership is modelled on security cooperation between GCHQ and the National Security Agency.
“This agreement marks a landmark moment as the UK and the US deepen our enduring special relationship to tackle the defining technological challenges of our generation,” Donnellan said. “Only by working together can we tackle technology's risks head on and harness its enormous potential to make lives better, healthier for all of us.”
The partnership follows commitments made at the AI Safety Summit held in the UK last November. The two agencies plan to develop a common approach to AI safety testing and share capabilities to effectively address risks. They will conduct at least one joint public testing exercise on openly accessible AI models and explore personnel exchanges.
Raimondo emphasized the importance of this collaboration, saying, “AI is the defining technology of our generation. This partnership will accelerate our laboratories' efforts to address the full range of risks to national security and society as a whole.”
Both governments recognize the rapid development of AI and the urgent need for a common global approach to its safety that can address emerging risks. This partnership is effective immediately and will enable seamless collaboration between our organizations.
“Working together, we will further strengthen the longstanding special relationship between the United States and the UK and lay the foundations for keeping AI safe, now and in the future,” Raimondo added.
In addition to joint testing and capability sharing, the UK and US will exchange key information on AI model capabilities, risks and underlying technical research, with the aim of strengthening a common scientific basis for AI safety testing that can be adopted by researchers around the world.
Despite the focus on risks, Donnellan argued that the UK has no plans to regulate AI more broadly in the short term. By contrast, President Joe Biden has taken a tougher stance against AI models that threaten national security, and EU AI law has introduced tougher restrictions.
Industry experts welcomed the collaboration, which they say is essential to fostering trust and safety in the development and deployment of AI across sectors such as marketing, finance and customer service.
“Ensuring that the development and use of AI is based on trust and security is paramount,” said Zoho's Ramprakash Ramamoorthy. “Having safeguards in place to protect training data reduces risk and increases confidence in those adopting AI solutions.”
Dr Henry Barani of Encompass added: “Mitigating AI risks through this collaboration agreement with the US is an important step towards reducing the risk of financial crime, fostering collaboration and supporting innovation in a technology sector where important advances are being made.”
(Photo: Art Lasovsky)
reference: IPPR: 8 million UK careers at risk of 'job destruction' due to AI
Want to learn more about AI and big data from industry leaders? Check out the AI & Big Data Expo in Amsterdam, California and London, a comprehensive event taking place alongside other major events like BlockX, Digital Transformation Week and Cyber Security & Cloud Expo.
Find out about upcoming enterprise technology events and webinars hosted by TechForge here.