# Altman warns AI could escape human control > Source: > Published: 2026-09-16 10:44:00+00:00 # Altman warns AI could escape human control [OpenAI](https://qazinform.com/tag/openai) CEO Sam Altman has warned that the public is justified in fearing the growing power of [artificial intelligence](https://qazinform.com/tag/artificial-intelligence) companies, while expressing confidence that the industry can develop the technology safely, reports a [Qazinform](https://qazinform.com/) News Agency correspondent. Speaking with Salesforce CEO Marc Benioff at the [Dreamforce conference](https://www.salesforce.com/dreamforce/) in San Francisco on Tuesday, Altman said some [AI](https://qazinform.com/tag/ai) developers could acquire excessive economic and social influence. ``` “I think the world is right to be afraid of this,” Altman said, warning that powerful companies could “exert undue influence on the economy” and impose particular worldviews on the public. ``` He nevertheless argued that [AI](https://qazinform.com/tag/ai) developers should be trusted to act responsibly. ``` “The world should trust that we are going to do the right thing because it’s the right thing and because we feel the magnitude of this,” he said. ``` Altman added that companies must keep their monitoring, alignment and safety systems ahead of rapidly advancing [AI](https://qazinform.com/tag/ai) capabilities. If that becomes impossible, they should be prepared to “slow down or stop.” In a separate [post](https://x.com/sama/status/2099352016988614852) on X, Altman identified two major dangers. ``` “First, we could lose control of the future to AI. This is unacceptable; we are unapologetically on Team Humanity, and AI must always serve people,” he wrote. ``` The second risk, he said, was an excessive concentration of power, potentially allowing a person, company, laboratory or country to impose its worldview on others. Nvidia CEO Jensen Huang offered a different perspective at Dreamforce, arguing that [AI](https://qazinform.com/tag/ai) safety was primarily “an engineering problem” rather than a matter requiring new legislation. ``` “Run as fast as you can, but if at any time you feel the institution is not in control, take a pause,” Huang said. ``` Earlier, [Qazinform](https://qazinform.com/) News Agency [reported](https://qazinform.com/news/openai-says-ai-cracked-century-old-maths-problem-511a64) that [OpenAI](https://qazinform.com/tag/openai) said it had [solved](https://openai.com/index/navier-stokes-solution/) a decades-old mathematical problem in just 88 hours using a new AI model and approximately 10,000 AI agents.