AI Safety Takes Center Stage as Tech Leaders Clash at Dreamforce

Artificial intelligence has rapidly advanced in recent years, with major tech firms racing to develop ever more capable models. This pace has sparked growing worries about how to keep AI safe and under control.

Concerns intensified after incidents where AI systems behaved unpredictably, prompting calls from some leaders for stricter oversight and slower progress. In September 2026, the debate reached a new level at Salesforce's Dreamforce conference in San Francisco, where executives from Nvidia, Anthropic, and OpenAI gathered to discuss the future of AI.

On September 14 and 15, 2026, Dario Amodei, CEO of Anthropic, published an essay and appeared in interviews urging the AI industry to slow down development. He argued that moving too fast could outpace safety measures and called for an 'AI kill switch' and global standards, according to The Hill.

At Dreamforce, Jensen Huang, CEO of Nvidia, took a different approach. While he agreed that unsafe AI models should not be released, he did not back a general slowdown of the industry's progress, CNBC International reported. Huang emphasized that companies should pause only if they cannot guarantee safety.

OpenAI's Sam Altman acknowledged the public's fears, saying the world is right to be concerned about the power concentrated among a few AI firms. However, Altman urged people to trust companies like his to manage the risks responsibly, as covered by the BBC and The Next Web.

The debate broadened as other tech leaders, including Meta's Mark Zuckerberg and Microsoft's Mustafa Suleyman, weighed in. Some criticized the idea of AI companies colluding to slow down progress, while others warned about the risks of AI systems being trained to mimic human consciousness.

The AI safety debate remains unresolved. Anthropic and OpenAI have proposed using internal AI risk evaluators, but critics question whether these tools will be powerful enough to prevent disasters.

Industry leaders are divided, with some pushing for more regulation and transparency, while others warn that too much coordination among firms could stifle innovation or even create new risks. For now, the world's most powerful AI companies are promising to act responsibly, but there is little agreement on what that means in practice.

The next steps will likely involve more public debate, possible regulatory proposals, and technical work on AI safety mechanisms. Whether tech leaders can agree on global standards or effective oversight remains an open question, and the pace of AI progress shows no sign of slowing.