Microsoft President Brad Smith emphasized the urgent need for global regulations on artificial intelligence (AI) in a recent interview, pointing out that the safety standards for AI are currently less stringent than those for basic products like milk.
In the BBC interview published Monday, Smith advocated for establishing a comprehensive framework of AI safety regulations. This includes mechanisms for notifying individuals when AI systems breach security, as well as an “emergency brake” that developers could activate. He suggested that there ought to be a legal obligation for companies to report incidents when AI systems fail, noting, “We already know how to handle this for other products. We should leverage that experience and think about how to apply it to AI.”
He further remarked, “At this moment, your washing machine is subject to more safety regulations than AI. Every carton of milk on grocery store shelves is inspected and regulated.” This stark contrast highlights the lack of oversight in the rapidly evolving field of AI.
As Smith speaks out, leaders from top AI companies like OpenAI and Anthropic, including CEOs Sam Altman and Dario Amodei, are also calling for a careful pace in the development of advanced AI technologies, stressing the importance of maintaining control over these systems.
Both Altman and Amodei have proposed that third-party evaluators be given similar access to AI model development processes as employees, allowing them to monitor for capability breaches and make such information public for wider understanding.
Smith believes that the UK is positioned to take the lead in AI regulation, noting the establishment of the AI Security Institute by former Prime Minister Rishi Sunak. He pointed out that the perspective in Washington differs significantly from the European Union’s regulatory measures, which categorize AI governance based on risk levels.
The AI Security Institute in Britain has the advantage of pre-deployment access to AI models, which is facilitated through voluntary agreements with key developers. This enables them to evaluate both the capabilities of AI systems and the potential risks they may pose. AI Minister Kanishka Narayan has indicated that Britain is open to introducing regulations for advanced AI models if the existing system does not suffice.
