Restricting the more capable variant to vetted organisations has a regulatory shape already. The General-Purpose AI Code of Practice, published 10 July 2025 and endorsed by the European Commission and the AI Board as an adequate voluntary tool, runs to three chapters, and its Safety and Security chapter applies only to providers of the most advanced models subject to systemic-risk obligations. My view is that gated access is the weakest of the available controls, because it moves the decision into the vendor's own vetting process and leaves no external record of who was refused. The Code does not ask for one.
Restricting the more capable variant to vetted organisations has a regulatory shape already. The General-Purpose AI Code of Practice, published 10 July 2025 and endorsed by the European Commission and the AI Board as an adequate voluntary tool, runs to three chapters, and its Safety and Security chapter applies only to providers of the most advanced models subject to systemic-risk obligations. My view is that gated access is the weakest of the available controls, because it moves the decision into the vendor's own vetting process and leaves no external record of who was refused. The Code does not ask for one.
That is an excellent point and I agree. To me right now it seems to be the best of many bad viable options as a 1st step
I wish I could use Mythos., but I have to say, Fable 5.1 is pretty powerful...