Michelle Donelan, the UK’s tech minister, emphasized the importance of international cooperation: “Ensuring the safe development of AI is a shared global issue.” The agreement builds upon commitments made at the 2023 AI Safety Summit, where both countries established AI Safety Institutes to evaluate various AI systems.

While the AI industry has seen rapid advancements, particularly in large language models like ChatGPT, Gemini, and Claude, regulations haven’t caught up yet. These companies haven’t faced significant restrictions, and haven’t been required to share details about their training data or the environmental impact of running their systems.

However, the EU’s recently adopted AI Act is expected to change this landscape. This legislation will require developers to be transparent about risks and share data used in certain AI systems. This transparency is crucial, as evidenced by OpenAI’s recent decision to shelve a voice cloning tool due to potential misuse, particularly during elections.

Currently, most AI systems perform specific tasks, often replicating human capabilities. These “narrow AI” tools raise less concern than the hypothetical “general AI” capable of broader human-like tasks. While some experts, like Professor Sir Nigel Shadbolt, believe fears about existential threats from AI are exaggerated, they acknowledge the need for caution and research on potential risks.

The US-UK agreement signifies a proactive approach to addressing AI safety concerns. As Gina Raimondo, the US commerce secretary, stated, the partnership aims to gain a deeper understanding of AI systems to provide better guidance and mitigate risks to national security and society as a whole. This collaborative effort sets a positive example for other countries to follow in ensuring the responsible development and deployment of AI for a brighter future.