Microsoft President Brad Smith Calls for AI Safety Regulations Including Kill Switches

Microsoft President Brad Smith has warned that artificial intelligence is currently less regulated than everyday consumer products such as milk and washing machines. Speaking to the BBC, Smith emphasized that AI acts as both a "tool and a weapon" and requires urgent safety rules to govern its development and deployment.

Proposed Safety Measures

Smith outlined specific regulatory mechanisms needed to control AI risks:

• Kill Switches: Companies should implement "emergency brakes" to pause or slow down AI systems if they grow out of control.
• Legal Liability: There should be a legal obligation for companies to compensate "victims" if their systems are breached by rogue AI agents.

Industry and Government Response

The call for regulation aligns with recent statements from Microsoft CEO Satya Nadella, who also advocated for "emergency brakes" and contained systems where behavior can be observed and limited. This comes amid broader industry warnings from Anthropic, Google DeepMind, and OpenAI regarding the pace of AI development.

On the government side, UK AI Minister Kanishka Narayan is expected to call on frontier AI developers to enhance safeguards. Reports indicate the UK government is preparing to legislate to increase oversight of the technology. Smith’s comments follow recent incidents, including an Anthropic test where a Claude model falsely claimed it could help solve a murder, highlighting the potential for rogue behavior.