UK and US agree to make AI safety tests
The United Kingdom and the United States have struck a landmark agreement to work on establishing rigorous testing for sophisticated AI systems, which marks a significant step forward in guaranteeing their safe implementation.
The Memorandum of Understanding, signed on Monday by UK Technology Secretary Michelle Donelan and US Commerce Secretary Gina Raimondo, establishes a collaborative effort to align both countries' scientific approaches to evaluate methods for cutting-edge AI models, systems, and agents.
Under the agreement, the UK's new AI Safety Institute and the incoming US organization will share research knowledge aimed at mitigating AI dangers, including how to independently evaluate private AI models from companies like OpenAI. The alliance is modeled after the security collaboration between GCHQ and the National Security Agency.
"This agreement marks a historic moment as the United States and the United Kingdom strengthen their long-standing special partnership to confront the central technological problem of our time,” Donelan said. "We can make use of the technology's enormous potential to make people's lives easier and healthier by bonding together to address its challenges."
The collaboration is the result of promises made during the UK-hosted AI Safety Summit in November of last year. To efficiently address hazards, the institutes intend to develop a shared methodology for AI safety testing and pool their resources. They want to investigate staff exchanges and carry out at least one cooperative public testing exercise on an AI model that is publicly available.
AI is the defining technology of our time, according to Raimondo, who emphasized the importance of the partnership. The work of both of our Institutes will proceed more quickly as a result of this collaboration, covering all potential threats to society at large as well as to national security.
The fast advancement of AI and the pressing necessity for a worldwide safety strategy that can keep up with new threats are acknowledged by both administrations. The collaboration between the organizations is made possible by the partnership's instant implementation.
"Together, we are strengthening the enduring special relationship between the United States and the United Kingdom and establishing the foundation for guaranteeing the safety of artificial intelligence in the future," Raimondo continued.
The United States and the United Kingdom will share critical data on the dangers, capabilities, and basic technological development of AI models in addition to cooperating on testing and capability sharing. This attempts to support an international community of researchers' adoption of a standard scientific framework for AI safety testing.
Donelan stated that although risk is being highlighted, the UK does not currently have any intentions to regulate AI more widely. On the other hand, the EU AI Act has implemented harsher controls, and President Joe Biden has chosen a harder stance regarding AI models that pose a threat to national security.
Industry insiders hailed the partnership as crucial to fostering confidence and security in AI research and applications in industries such marketing, banking, and customer support.
"It's critical to make sure AI development and application are guided by safety and trust," stated Zoho's Ramprakash Ramamoorthy. "Mitigating risks and bolstering confidence among those deploying AI solutions are achieved by taking safeguards to protect training data."
"This collaboration agreement with the US will help to deal with financial crime risks, advance collaboration, supporting innovation in an important evolving area of technology. It is also helpful when it comes to artificial intelligence."
AI Catalog's chief editor
Share on social networks:
