• The U.S. and U.K. have entered into a pioneering bilateral agreement to jointly develop safety tests for evaluating AI systems.
  • The Memorandum of Understanding signed by U.S. Commerce Secretary Gina Raimondo and U.K. Technology Secretary Michelle Donelan facilitates information and technical knowledge exchange on AI model risks and security measures.
  • Dedicated AI Safety Institutes from both countries will collaborate under this agreement to establish safety testing measures and guidelines, aiming to address national security concerns and broader societal impacts associated with AI.

The United States and the United Kingdom have committed to collaboratively develop tests to assess the safety of AI systems, marking a groundbreaking bilateral agreement.

Bilateral agreement to enhance AI safety testing

Commerce Secretary Gina Raimondo of the U.S. and Technology Secretary Michelle Donelan of the U.K. have signed a Memorandum of Understanding aimed at facilitating the exchange of information and technical expertise concerning the risks associated with AI models and their security measures between the two nations.

Also read: Microsoft safety system can catch hallucinations in its AI apps

This announcement follows the introduction of new requirements for U.S. federal agencies regarding the responsible utilization of AI.

Under this agreement, the dedicated AI Safety Institutes of both nations will cooperate in establishing measures for safety testing.

Raimondo emphasized, “This partnership will expedite the efforts of our respective Institutes across a comprehensive range of risks, whether pertaining to our national security or wider society. Our collaboration underscores our commitment to confronting these concerns head-on rather than evading them.”

In the aftermath of last year’s AI Safety Summit, both the U.S. and the U.K. established AI safety institutes tasked with formulating guidelines and benchmarks for assessing AI risks.

Researchers at the U.S. AI Safety Institute (U.S. AISI) will collaborate with their U.K. counterparts to develop a unified approach to AI safety testing.

While it has been announced that joint testing exercises will be conducted on publicly accessible models, no specific timeframe has been provided for these activities.

Key highlights of the U.S. and U.K. collaboration

According to U.K. Tech Secretary Michelle Donelan, this agreement signifies a pivotal model for strengthening the bond between the two nations in tackling “the defining technology challenge of our generation.”

Donelan remarked, “We have consistently emphasized that ensuring the safe development of AI is a shared global concern. Only through collaborative efforts can we directly address the risks associated with this technology and leverage its immense potential to improve our lives.”

These tests are a direct outcome of the AI partnership between the U.S. and the U.K. and could offer valuable assistance to startups, as noted by Ekaterina Almasque, general partner at venture capital firm OpenOcean.

Almasque stated, “Startups in AI frequently encounter challenges navigating the intricate landscape of safety and ethics, such as acquiring ethical training data at a reasonable cost, which can hinder their ability to innovate, expand, and develop competitive products. This collaboration establishes a framework for addressing these obstacles, providing guidance and support to assist startups in responsibly and securely developing AI technologies.”