The United Kingdom and United States governments have announced a joint effort to enhance safety testing for advanced artificial intelligence (AI) models. In a move aimed at addressing the evolving landscape of AI technology, an agreement was signed by Michelle Donelan, the UK Secretary of State for Science, Innovation, and Technology, and US Secretary of Commerce Gina Raimondo.
Donelan described the agreement as a pivotal moment in the journey of AI safety, emphasizing the collaborative role of both nations in shaping a network of institutes dedicated to AI safety. The UK and US AI Safety Institutes, established around the inaugural AI Safety Summit at Bletchley Park in November 2023, are set to further formalize their cooperation through this agreement.
The agreement outlines plans for the development of a shared approach to AI safety testing, facilitating the exchange of personnel and information between the two institutes. Notably, joint testing exercises on publicly available AI models are on the agenda, signaling a concerted effort to address AI safety challenges on a global scale.
Secretary Raimondo emphasized the shared responsibility of ensuring safe AI development, highlighting the importance of international collaboration in addressing the capabilities and risks associated with AI models and systems.
Such safety tests are gaining prominence in the context of AI advancement, with tech companies and lawmakers increasingly focusing on mitigating potential risks. Notable companies like OpenAI and Anthropic have outlined strategies to incorporate safety tests into their product development processes. Legislative initiatives, such as the recent E.U. AI Act and President Biden’s executive order on AI, underscore the growing regulatory scrutiny of powerful AI models.
The UK has been a leading advocate for international action on AI safety, hosting the inaugural AI Safety Summit and committing significant funding to the UK AI Safety Institute. While the US has also shown commitment, the disparity in financial contributions has raised questions. However, Donelan emphasized the broader expertise and resources available within the US government beyond direct funding.
Despite their leadership roles, both the UK and US governments have refrained from immediate regulatory measures on AI. While some advocate for mandatory disclosure of AI safety test results, Donelan emphasized the importance of fostering innovation without stifling industry growth.
The memorandum between the UK and US also paves the way for similar partnerships with other nations, reflecting the global nature of AI challenges and the need for collaborative solutions. Donelan stressed the imperative of international cooperation in harnessing the potential of AI for the collective benefit of humanity.