The Fable 5 "safety cage" is doing a lot of PR work and nobody's talking about it

Anthropic spent last week publicly warning that frontier AI is getting dangerous enough to need a coordinated industry slowdown. Three days later they shipped Fable 5 — their own words: the most capable model they've ever released to the public. The trick: Fable 5 and the restricted Mythos 5 are the same model. The only difference is classifiers in front that silently reroute "dangerous" queries to an older model. The model isn't safe — the cage is. And the cage is the entire reason they get to sell it. Oh, and to ship it they imposed 30-day data retention on all traffic, including enterprises that had zero-retention agreements. Your privacy guarantee got downgraded so the launch could happen. There's also an IPO coming. Maybe the cage genuinely works — no universal jailbreaks in 1000+ hours of red-teaming is legitimately impressive. But "too dangerous to slow down for, safe enough to sell" is a hell of a needle to thread. Am I being too cynical or does the safety story not survive the release notes?