Anthropic CEO Dario Amodei is pushing to embed outside evaluators into AI architecture to police what the technology can say — and the watchdog he has in mind is run by his own former colleagues and funded by the same network that bankrolled his company.
This is the censorship pipeline upgrading from deplatforming to pre-publication thought control. If evaluators are baked into AI architecture, a handful of Silicon Valley executives and their hand-picked auditors decide which questions Americans can ask and which answers they're allowed to receive — before a word ever appears on screen.
Amodei has called for independent outside watchdogs to safeguard AI development. But Fox News reported that one organization he backs for the role, METR, is tightly connected to the same AI safety community that spawned Anthropic itself. METR describes itself as an AI safety testing laboratory. Its founder and CEO, Beth Barnes, worked at OpenAI alongside Amodei. Its earlier founder, Paul Christiano, led OpenAI's alignment research. Both have framed their work through the lens of
