Microsoft's Brad Smith Calls for Stricter AI Regulation

Microsoft vice chair and president Brad Smith has emphasized the need for stricter regulation of artificial intelligence, stating that it is significantly under-regulated. In an interview with the BBC, he noted that current safety rules are more stringent for household appliances like washing machines than for AI systems.

In fact, Smith pointed out that even cartons of milk in grocery stores are subject to regular inspections and adherence to strict safety standards. He believes this highlights a glaring disparity between the level of oversight applied to everyday consumer products versus cutting-edge technologies like AI.

AI, according to Smith, is both a valuable tool and a potential threat when it malfunctions or falls into malicious hands. To mitigate these risks, he advocates for implementing ‘kill switches’ that can be activated in emergency situations. This would allow organizations to quickly shut down compromised systems and prevent further damage.

In addition to kill switches, Smith recommends establishing clear incident reporting requirements for companies whose AI security is breached. He also suggests that victims of such incidents should be promptly informed about the nature of the breach and any steps being taken to rectify it.

Smith’s call for stricter regulation has been echoed by Microsoft chief executive Satya Nadella in a recent blog post. Nadella emphasizes the need for trust architecture in AI systems, arguing that relying solely on model providers’ assurances is insufficient. He advocates for principles of observability, verifiability, and independent auditability to ensure accountability.

The concept of ‘Super Intelligence,’ popularized by Donald Trump’s preferred term, has been referenced by Nadella as a key aspect of the current AI landscape. However, this label has also sparked controversy, with some critics arguing that it downplays the complexity and nuance of AI systems.

Despite these challenges, Smith remains optimistic about the UK’s potential to lead global efforts in AI safety regulation. He notes that the country is well-positioned due to its strong technical capabilities, as evidenced by the work of organizations like the AI Security Institute (AISI).

In a recent development, Microsoft has joined several prominent tech companies in signing a pledge to collaborate with independent auditors on their AI systems. However, this move has been met with skepticism by some critics who view it as merely ‘morally binding’ rather than legally enforceable.

Trump’s stance on AI regulation remains ambiguous, having previously expressed reluctance to impose strict controls on the industry. His recent call for AI developers to refrain from sharing unreleased systems with AISI led to a temporary ban on UK officials examining Anthropic’s Claude Mythos 5.1 model. However, this restriction has since been lifted.

Smith believes that a strong stance on AI safety regulation could be viewed positively in the US and help bring people together around this critical issue. He emphasizes the need for calm and collaborative discussions to address the challenges posed by rapidly advancing AI technologies.

The UK’s leadership role in AI safety efforts is seen as a potential strength, with Smith noting that Washington DC views London differently than Brussels. This distinction could facilitate greater international cooperation on regulating AI systems.