Microsoft President Brad Smith has publicly urged governments to implement stricter safety regulations for artificial intelligence, arguing the technology currently faces fewer safety rules than consumer goods like milk or washing machines. Speaking to the BBC, Smith advocated for urgent legal frameworks, including statutory liability to protect victims if systems are breached by rogue AI agents, alongside mandatory emergency brakes or kill switches.

Echoing Smith’s comments, Microsoft Chief Executive Satya Nadella called for contained systems featuring observable behaviors, testable limits, and absolute human containment options. Nadella stated that superintelligence cannot be treated as a series of nested black boxes. Their warnings coincide with statements from frontier labs—including Anthropic, Google DeepMind, and OpenAI—acknowledging that AI development rates may need to be managed carefully to preserve control.

The regulatory pressure comes as the UK government prepares to introduce legislation expanding oversight for frontier AI models. UK AI Minister Kanishka Narayan is expected to demand increased risk mitigation from AI developers, while cybersecurity experts warn that unreleased frontier models remain vulnerable to theft by state adversaries.

Why it matters

  • Enterprise software operators should expect regulatory mandates forcing predictable safety controls and auditability into AI agents.

  • Third-party liability for autonomous agent actions may soon become a statutory reality for AI deployment teams.

  • Kill switches and containment architecture will become required features in enterprise agent deployments.

Source: theguardian.com