Altman reportedly outlines three paths for superintelligence oversight

Sam Altman reportedly discussed three possible approaches to managing superintelligence during an internal conversation with OpenAI employees. The unconfirmed account describes alternatives for governing advanced AI, including one company gaining enough power to prevent rivals from competing..

The first approach would establish an international safety authority modeled on the International Atomic Energy Agency. A second would combine leading AI laboratories into a shared research organization. Altman reportedly favored international oversight and criticized the third scenario, in which a developer uses self-improving AI to gain overwhelming power over competitors’ computing resources..

International oversight already has a place in OpenAI’s published policy proposals. In May 2023, Altman, Greg Brockman and Ilya Sutskever proposed an authority empowered to inspect advanced systems, require safety checks and restrict deployment. They also discussed coordinating leading development efforts, including through a project organized by governments..

Altman reportedly told employees that OpenAI should decide within the following year which approach to pursue. That remains an unconfirmed account of an internal discussion, rather than an announced implementation schedule. The reported remarks do not establish that OpenAI has adopted a plan to destroy competing AI hardware..