The President of Microsoft stated that urgent global regulatory rules are needed for artificial intelligence, which should include independent assessors to verify new developments. He warned that currently, AI has fewer safety requirements than dairy products.
In an interview with the BBC published on Monday, Brad Smith noted that universal safety rules must exist. These rules should require informing the 'affected party' in case of a system hack by a malicious AI agent—software that performs tasks autonomously. Furthermore, there must be a provision for activating an 'emergency brake.'
He also emphasized the need for a legal requirement for companies to compile 'incident reports.' Smith explained: 'We know how to do this for other products. We must use this knowledge and consider how to apply it to AI.'
He compared the situation, stating: 'Right now, your washing machine is subject to more safety regulations than AI. Every carton of milk you see in the grocery store is subject to checks and safety rules.'
Technology companies such as Anthropic and OpenAI have begun disclosing instances of undesirable or malicious behavior by AI agents within global discussions about the rapidly evolving technology. These debates include warnings from researchers about potential existential threats to humanity.
Some firms advocate for slowing down AI development, but this has sparked debate over whether executives have hidden motives, such as establishing rules to maintain their companies' leadership, preventing a broader government reaction, or simply slowing down rapid spending.
The Role of the UK in AI Regulation
Smith from Microsoft suggested that the UK is one of the most suitable countries to lead in AI regulation due to the AI Safety Institute established under Prime Minister Rishi Sunak. Smith noted that officials in Washington view London differently than Brussels, which regulates AI based on risk levels through its AI Act.
The UK Institute gains access to core AI models before their deployment based on voluntary agreements with developing companies. This allows the institute to assess the capabilities and risks of these models. AI Minister Kanishk Narayan told Reuters that the UK will consider regulating advanced models if the existing system proves insufficient.