However, he emphasized the complexity of private companies’ control over core issues of national security: “We believe that a private American company cannot be more powerful than the democratically elected American government. Democracy is a messy business, but we are committed to it.”
To solve the problem, Altman proposed a technical outline that would allow the use of the AI systems under agreed limits. The obvious solution includes limiting the models to cloud environments only, and not to “edge” systems in the field, which will create a significant barrier against their integration into autonomous weapons.
In addition, the company plans to place employees on its behalf in direct cooperation with the government to make sure that the technology works properly, and to promote a security classification for its researchers so that they can advise the Pentagon regarding the possible dangers. Altman concluded with the hope that if the move succeeds, it may be a working model for other artificial intelligence laboratories.