Every AI can act in 2026. The Rad family is the one you can trust on the consequential stuff — because the dangerous moves are refused in code, not by the model's mood.
A tool, not a guarantee. One valve, every surface.
one line runs through everything — it glides mint where the move is safe, and holds amber at the line a human owns.
One valve, every surface. Money, changed banking, first-contact, decisions, prompt-injection — and a robot's irreversible reach — are human-only, forever, decided deterministically.
No prompt talks any Rad product into doing the thing it shouldn't. The refusal doesn't depend on which model is running, how it's feeling, or how clever the attacker is. It's refused in code.
Five surfaces. One refusal you can trust.
The same promise runs through all of them: the consequential action is human-only, forever.