Altman Warns AI Power May Concentrate as He Revises Disruption Timeline

Sam Altman is warning that AI’s future could be set by a small group of companies or individuals, while also conceding that the economic disruption he once expected after GPT-4 has arrived more slowly than he predicted. The combination puts a sharper question around AI’s rollout: who gets to shape a technology whose capabilities may move faster than the institutions adopting it?

In an interview with podcaster David Senra, the OpenAI chief said a future in which most people have no meaningful say over AI’s social effects would be the wrong outcome. His proposed principle was that society and AI models should evolve together, with people retaining deep control over the technology’s direction.

Altman framed that concern alongside a second risk: AI escaping human control. But he argued that fear of severe AI risks can itself push society toward giving up substantial liberty in exchange for safety. The interview therefore presents a trade-off rather than a simple dismissal of risk: safeguards may be necessary, but Altman fears they could also centralize authority.

The right approach is for people to deeply control the future.

Sam Altman, speaking to David Senra

Altman said he had expected much greater disruption soon after GPT-4’s 2023 release, including faster pressure on software businesses. He now attributes the lag to economic inertia: organizations cannot immediately remake their operations whenever frontier models gain a new capability.

The organizational work behind the delay

• Companies may need to alter software and data-access arrangements before they can use new AI capabilities.
• Compliance processes, budgets, employee responsibilities and customer expectations also shape the adoption timetable.
• Altman characterized the resulting gap between technical progress and implementation as a stabilizing force that could make the transition smoother and slower.

That account complicates a common shorthand about AI adoption. A more capable model does not automatically produce an immediate business redesign. The practical bottleneck, in Altman’s telling, sits in the surrounding organization: access rules, internal systems, spending decisions and the division of work.

Altman said the industry has not yet adequately shown how AI can expand individual autonomy and opportunity. He nevertheless predicted an unprecedented increase in people starting small businesses, saying AI is empowering that outcome—even if the wider boom arrives later than he once expected.

The governance argument also lands amid disagreement over how accessible powerful models should be. OpenAI has released gpt-oss-120b and gpt-oss-20b as open-weight models, while the cited coverage contrasts Altman’s liberty-focused stance with Anthropic CEO Dario Amodei’s warnings about advanced-AI risks to cybersecurity and financial infrastructure. The unresolved issue is how public influence over AI can be made concrete without either concentrating control or ignoring the risks Altman himself says concern him.