OpenAI, Google, Anthropic, and other companies developing generative AI are constantly improving their technology and releasing better and better large-scale language models. The UK and US governments have signed a memorandum of understanding to develop a common approach to independent assessment of the safety of these models. The UK’s AI Security Institute and the US’s AI Security Institute (announced by Vice President Kamala Harris but not yet operational) will jointly develop a suite of tests to assess risks and ensure “the security of state-of-the-art AI models” “. “
They plan to share technical knowledge, information and even personnel as part of the partnership, and one of their initial goals appears to be joint testing on publicly accessible models.British Science Minister Michelle Donelan, who signed the agreement, said financial times They “really have to move quickly” as they expect a new generation of AI models to be available next year. They believe these models could be a “complete game changer,” but they still don’t know what they can do.
according to era, While both the US and UK intend to work with other countries in the future, this partnership is the world’s first bilateral arrangement on AI security. “Artificial intelligence is the defining technology of our generation,” said U.S. Commerce Secretary Gina Raimondo. “This partnership will accelerate the work of our two institutes to address the full range of risks, whether they are to our country or not.” safety or our wider society.” “Our partnership makes it clear that we are not escaping these concerns – we are working hard to address them. As a result of our collaboration, our institutes will gain a better understanding of artificial intelligence systems , conduct a more robust assessment and issue stricter guidance. ”
While this particular partnership focuses on testing and evaluation, governments around the world are also developing regulations to control AI tools. Back in March, the White House signed an executive order aimed at ensuring that federal agencies only use artificial intelligence tools that “do not jeopardize the rights and safety of the American people.” A few weeks ago, the European Parliament approved sweeping legislation regulating artificial intelligence. It would ban “artificial intelligence that manipulates human behavior or exploits human vulnerabilities”, “biometric classification systems based on sensitive characteristics”, and “untargeted scraping” of faces from CCTV footage and the Internet to create facial recognition databases . Additionally, deepfakes and other AI-generated images, videos and audio will need to be clearly labeled according to its rules.
