Mira Murati is pushing back against the idea that more capable AI must automatically mean less human control.
In comments highlighted by WIRED, the former OpenAI CTO argued for systems that keep people actively involved in important decisions rather than treating full autonomy as the inevitable destination of model development.
The human-in-the-loop framework is increasingly central to debates about how AI should be deployed in software, medicine, research, education, and enterprise operations. Supporters see it as a way to combine model speed and pattern recognition with human context, responsibility, and judgment.
Murati's position also reflects growing skepticism toward marketing narratives that describe AI systems as stand-alone thinking machines. Even when models become more capable, reliability, context awareness, and alignment with human intent remain uneven.
That distinction matters because the consequences of over-automation are not abstract. Poorly supervised systems can amplify mistakes, obscure accountability, and encourage organizations to remove human review from the places where it matters most.
The discussion lands at a time when many technology companies are racing to build agents that can use tools, execute workflows, and take multi-step actions with limited supervision. The commercial pressure to make those systems feel autonomous is high, but so is the risk of overstating what they can safely do.
Murati's stance therefore matters beyond one interview. The norms that product leaders establish now will shape whether advanced AI becomes an assistive layer that augments human operators or a control layer that quietly displaces them from consequential choices.