Microsoft President Brad Smith on Tuesday called on industry and governments to act without delay to create real and effective safety standards to contain the risks associated with artificial intelligence (AI). He said this to journalists on the eve of the international AI summit in Dublin.
"We will need for AI the same thing we have for everything else in life... we need real and effective safety standards, and we need the public to trust them," Smith said.
Concern about AI safety has risen sharply in recent months after models developed by OpenAI and Anthropic were involved in safety incidents during testing. This has increased pressure for regulation of the technology.
"We have reached a point where industry companies clearly need to step up and do more to address AI safety problems," Smith said. "I think the time has come when governments also need to do more."
The Microsoft head set out the company's proposals for strengthening safety in the industry. The first is a requirement to equip AI agents (programs capable of autonomously performing tasks) with safety brakes that users could activate in the event of an incident. "The idea is that, especially for any AI operating on critical infrastructure or another automated system, there should be the ability to turn it off or slow it down," he explained.
Smith also called for mandatory incident reporting requirements for AI providers. "If you are a business and you have suffered a leak that led to the loss of consumer data, you are obliged to notify consumers," he said. "The same logic should apply to people harmed by an AI agent," with a legal obligation to inform those affected.
In addition, the Microsoft president advocated stricter testing and evaluation of AI models before release to make sure they are safe. "We need to act without delay," he warned.
Smith is not the first leading industry figure to point to the need for heightened vigilance around AI. Among them is Anthropic head Dario Amodei, who has promised to unilaterally slow development if necessary. Nvidia CEO Jensen Huang said on Tuesday that leaders of AI companies must be responsible for the safe development of the technology. Speaking in Paris, he said: "The full potential of AI can be realized only if we advance AI capabilities and its safety at the same time. These are not conflicting goals. In fact, they are the same goal."