Sam Altman and Greg Brockman say people should remain in charge of AI’s future, while warning that current ways of monitoring advanced models may become less effective. Their comments at the New York Times DealBook Summit put a practical question behind the familiar principle: how do you keep people in control as models become more capable?
Watch Desk analysis
What happened
Altman said a gradual handover of decision-making authority to AI would be “a terrible idea”, according to TechCrunch’s account of the summit. He said he and Brockman agree on the goal of keeping people in control, though Altman described them as differing on how to achieve it.
Brockman also said that monitoring may need new approaches. Researchers currently monitor models’ chain of thought, but he argued that more capable models may do more without spelling out their reasoning, weakening that method.
Why it matters
This is more than a declaration of principle. If a model’s written reasoning becomes a less dependable window into what it is doing, developers will need other ways to spot problems and assess behaviour. Brockman said new monitoring technology will be needed; the account does not set out what that technology would be or when it might arrive.
The remarks also expose a tension at the heart of AI development: building more capable systems while ensuring people can meaningfully oversee them. “Keep humans in control” is a fine destination. The engineering route matters rather more than the slogan.
Our read
The useful part is Brockman’s specific warning about chain-of-thought monitoring, not another round of agreement that safety matters. It points to a technical challenge, while leaving the proposed answer open. Readers should treat these as summit remarks, not evidence that a workable replacement monitoring system is already in place.
What to watch
- Whether OpenAI describes monitoring methods beyond models’ written reasoning.
- How the company evaluates oversight as models become more capable.
- Whether Altman and Brockman explain where their approaches to keeping people in control differ.
Discussion spark: Should AI developers rely on models’ written reasoning as an oversight tool, or should they be required to demonstrate other ways of monitoring them?
Sources and evidence
- OpenAI Executives Say They Have to Strike a Balance Between A.I. Safety and Progress – The New York Times (18 November 2026, 17:03 UTC)
- Sam Altman warns against a future where AI, not humans, calls the shots – TechCrunch (18 November 2026, 17:03 UTC)
- OpenAI CEO Sam Altman says it’s ‘hard to imagine the AI industry without Jensen’ – TechCrunch (18 November 2026, 17:03 UTC)
Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.