Sam Altman says cognitive atrophy is one of the most overlooked risks of the AI era

OpenAI CEO Sam Altman Discusses Potential Effects of AI on Human Cognitive Skills

OpenAI chief executive Sam Altman has identified reduced mental effort as one area of AI development that receives limited public discussion. He raised the possibility that widespread use of the technology could lead to weaker individual reasoning abilities if users delegate too many cognitive tasks to automated systems.

Altman made the remarks during an interview focused on open questions in artificial intelligence. He asked how society could ensure that AI tools encourage greater mental engagement rather than replace it. The comments addressed both current capabilities and the direction of future systems that handle more complex professional work.

Researchers have examined similar questions in controlled settings. Nataliya Kosmyna, an investigator at the Massachusetts Institute of Technology, conducted preliminary experiments in which participants wrote essays with the assistance of generative AI models. The study recorded lower performance on subsequent tasks compared with groups that used only search engines or no external aid. Kosmyna noted that no prior technology has allowed users to outsource extended sequences of reasoning in the same way.

Theoretical neuroscientist Vivienne Ming has used the term chronic substitution to describe repeated reliance on AI for tasks previously performed by individuals. Ming has stated that sustained substitution may reduce cognitive reserve, a measure associated with resilience against age-related decline such as dementia. These observations apply to tools that now perform document review, data analysis, customer-service responses, and software coding.

Paris-based company Mistral AI chief executive Arthur Mensch has expressed parallel concerns. In an interview last year, Mensch described the possibility that workers could experience progressive deskilling if they increasingly turn to AI systems for information retrieval and routine decisions. The pattern, he suggested, could reduce independent problem-solving over time.

Altman also outlined expected benefits from continued AI progress. He described systems that remain active throughout the day, summarizing documents, recording meeting content, retrieving earlier correspondence, and using overnight processing cycles to propose next steps. He characterized the overall outlook as potentially transformative for personal productivity.

At the same time, Altman indicated that certain human capabilities may remain difficult to replicate. He referred to forms of judgment that current models have not matched and suggested that new terminology may be required to distinguish these abilities from machine outputs.

The discussion occurs as AI systems are integrated into additional workplace functions. Organizations are testing models for research synthesis, legal document processing, and technical troubleshooting. Studies on long-term skill retention remain limited, and researchers continue to track whether performance gains from AI assistance persist when the tools are removed.

Altman’s statements align with earlier comments from other technology executives who have called for attention to both capability gains and possible reductions in practiced human skills. No single study has established the magnitude of any long-term effect across populations, and outcomes may vary by task type, frequency of use, and individual training.

Further work is underway to measure changes in reasoning speed, accuracy, and adaptability among users of different AI configurations. These investigations include both laboratory protocols and workplace observations. Results are expected to inform guidelines on appropriate levels of AI assistance in education and professional training programs.

The conversation reflects ongoing examination of how automated systems interact with established human cognitive processes rather than a settled conclusion about net outcomes.