MUMBAI: The AI race is accelerating, but Sam Altman says the brakes should still work when they are needed. Speaking at Salesforce’s annual conference in San Francisco on Tuesday, the OpenAI CEO said public concern over increasingly capable AI systems was justified, while arguing that companies developing the technology could take responsibility for keeping its risks in check.
Altman acknowledged that imagining serious consequences from advanced AI is becoming easier as the technology progresses. “It doesn’t take as much imagination as it used to for [us] to imagine how this could go wrong,” he said, adding that “the world is right to be afraid of this”.
At the same time, Altman said OpenAI and other AI companies should be trusted to act responsibly rather than relying entirely on government intervention. He said the industry has the ability to put safety measures in place as systems become more capable.
That confidence, however, came with a clear condition. Altman said OpenAI and other companies would need to slow down or stop development if they reached a point where they could no longer keep safety ahead of capabilities.
“The world should trust that we are going to do the right thing because it’s the right thing and we feel the magnitude of this,” Altman said.
His comments arrive as the AI industry’s safety debate gathers momentum following warnings from researchers and executives about the consequences of pushing frontier systems too quickly.
A researcher who recently left Anthropic warned that AI could pose an extreme threat to humanity by the end of the decade if development remained unchecked. The warning triggered wider discussion among AI executives and experts, with some expressing agreement with the assessment, although the detailed reasoning behind those views was not set out.
Anthropic CEO Dario Amodei has separately argued that the pace of AI development should be reduced and has called for governments to introduce regulation. Anthropic is also discussing common safety standards with other AI companies, according to the report.
OpenAI executive Chris Lehane said last week that OpenAI was working with other frontier AI labs, including Anthropic and Google DeepMind, on a voluntary effort to develop safety standards.
The question of who should hold the safety steering wheel has also drawn support for greater industry control from other technology leaders.
Meta CEO Mark Zuckerberg has argued that AI companies have both the ability and incentive to focus on alignment, pointing to competitive pressure and potential liability if systems cause harm. He said companies that fail to prioritise alignment could risk falling behind.
Nvidia CEO Jensen Huang has taken a similar position, saying decisions on whether new AI systems should be released should remain with the companies developing them. Huang described AI safety as an engineering challenge and argued that additional laws or regulations were not necessary.
Like Altman, Huang has also put a line in the sand: AI companies should pause development if their leaders lose confidence that they remain in control of their systems.
That leaves the industry’s central debate less about whether AI needs safety measures and more about how those measures should work and who gets to decide when the technology has moved too fast.
