Sam Altman Says Fear of AI Companies’ Power Is Justified

Sam Altman says public anxiety about AI companies amassing too much power is warranted. At a Salesforce Dreamforce event in San Francisco, the OpenAI chief said developers could gain undue influence over the economy and push a worldview onto people—risks he placed alongside the possibility of serious loss-of-control accidents.

The remarks, made September 15, are notable for their directness. Altman did not frame distrust of powerful AI developers as a misunderstanding to dismiss. He described it as a reasonable response to the prospect that a small number of companies could hold enough influence to shape economic outcomes or impose their values more broadly.

Two risks, one demand for confidence

Altman separated the challenge into two parts. One is a loss-of-control accident or another serious failure. The other is excessive concentration of power among the organizations developing AI. His account treats those as related pressures: society must guard against systems going wrong while also resisting a future in which their builders can exercise outsized influence.

• A serious accident in which AI development or deployment goes wrong.
• Power becoming concentrated enough for AI developers to influence the economy or press a worldview on others.

And I think the world is right to be afraid of this.

Sam Altman, speaking at Salesforce Dreamforce

The answer Altman offered

Altman’s response was not a specific new safeguard or an outside oversight proposal in these remarks. Instead, he called for pragmatism and steadiness, saying the public should trust OpenAI and other developers to make dependable, consistent decisions in light of the stakes.

That is a demanding position for an industry whose leaders are themselves central to the power-concentration concern Altman described. His argument asks people to accept that the danger is real while putting substantial weight on the judgment of the companies that could acquire that influence.

Trust is the unresolved test

Altman said there is a narrow path through the problem: decisions must be pragmatic, steady, dependable and consistent enough for the world to trust them. But the tension in his comments remains intact. Acknowledging that AI companies could gain undue economic and cultural influence does not by itself settle how that influence should be constrained, or what would make a company’s assurances credible to the people affected by its choices.