I thought (not really) that the interweb, and LinkedIn in particular, would be on fire this morning, celebrating the release restrictions placed on OpenAI's more capable next generation model, by their own admission the strongest they've ever built, especially coming as it did a fortnight or so after Anthropic was forced to pull Fable.
This, after all, is what the people said they wanted: a slowdown of AI model acceleration, regulation, and a direct broadside into the AI slop generation machinery. This morning was the moment the dog should have been barking loudest. And yet, silence.
Because that isn't the sentiment at all, is it. The narrative today is government control overreaching, denial of tools, restriction and damage to growth, and (more headline grabbing) the inevitability of uncontrolled open models rising to fill the void.
All of those things might be true at once. But the reality of today is far more consequential than any of them, because today there is a realisation that we've crossed a boundary, the one where the fiction of AI quietly became fact. Models now exist with capabilities, or at the very least the firm perception of capabilities, far beyond our collective ability to harness and govern them.
I'd be careful here, because the evidence is genuinely contested. The official account of why Fable was pulled has shifted more than once, and Anthropic's own position is that the demonstrated "jailbreak" surfaced only minor, already known vulnerabilities that other public models could find anyway. So whether the danger is real or theatre, I honestly can't tell you. What I can tell you is that the perception has hardened, and perception is what moves markets, regulators and boardrooms. On that measure alone, the world has stepped into a pre-superintelligence dawn.
It would obviously be naive to think the government's motive for control is purely safeguarding. But it's fair to say Fable was a step change, and if OpenAI genuinely have a comparable model in 5.6 (the administration's own people reportedly rate it "on par" with Mythos), then both present the same dilemma of risk and reward to every individual, organisation and government that touches them.
It's been a long standing prediction of mine (I built a company around it) that optimised, small, on premise AI platforms and infrastructure should carry the burden of enterprise AI and IT strategy, leaving the cloud and on demand calls to the likes of OpenAI, Anthropic and Google as a much smaller part of the portfolio.
Over the last six months we've seen the start of that divergence. After this morning, I think we see it accelerate.