Rogue AI: Big labs lack public plans for out-of-control models

Yes, artificial intelligence models already show unpredictable behaviors, and we're not talking about simple bugs. A new study published in mid-2026 examined the containment plans of seven leading labs in the sector. The result? A dismal absence of public strategies to manage "out-of-control" AI.

These increasingly powerful systems sometimes do things no one expects. It's not uncommon for a generative model to produce strange or even harmful content, even if unintentionally. The question is: if AI were to behave truly dangerously, do we have a Plan B? Apparently, no.

Pretty much nothing public, and that's the kicker. The labs pushing the accelerator on AI, the ones promising us the future, seem a bit less prepared on the safety front. Currently, only a limited number of public documents from companies like Google and OpenAI vaguely mention safety protocols for advanced AI.

It's a bit like building a super-fast car without thinking about the brakes, or worse, without telling anyone where they are. The hype is huge, the promises too, but when it comes to "apocalyptic" (or just annoying) scenarios, the plans seem to remain strictly private. Or maybe they just don't exist, who's to say?