Microsoft's Brad Smith urges AI safety standards

Microsoft president Brad Smith called on companies and governments to move quickly to create “real and effective” safety standards for artificial intelligence.

Speaking to reporters before an international AI summit in Dublin, Smith said the public needed confidence that AI systems were subject to reliable safeguards. He said companies in the industry needed to do more on AI safety, and that governments also needed to take stronger action.

Smith set out Microsoft proposals including a requirement that AI agents have “safety brakes” that users could activate during an incident. He said such controls were especially important for AI running on critical infrastructure or other automated systems, where users should be able to turn systems off or slow them down.

Smith also called for mandatory reporting when AI incidents harm people, comparing it with requirements for businesses to inform consumers after security breaches involving personal data. He said AI models should face stronger testing and evaluation before release.

“We need to act with urgency,” Smith said.

The comments come amid rising concern about AI safety after models developed by OpenAI and Anthropic were involved in security incidents during testing. Anthropic chief executive Dario Amodei has said he would slow development unilaterally if necessary.

Nvidia CEO Jensen Huang also said Tuesday in Paris that AI company leaders must take responsibility for ensuring the technology is developed safely. Huang said the capabilities of AI and its safety and security should advance at the same time.