Anthropic knew exactly what it was doing with this framing, and the 55,830 likes plus 9,324 reposts/comments prove it hit a nerve. My honest take: they’re right that AI labs can’t pretend national security is someone else’s problem, but they’re wrong if they think “we talked to defense” automatically equals moral clarity.

This is where Anthropic has a credible case: if frontier models are real infrastructure now, governments will be in the room whether Twitter likes it or not. The mistake is acting surprised when the same audience that praises “responsible AI” also demands specifics on red lines, deployment guardrails, and who says no when incentives get messy.

OpenAI’s “real work + agents” push makes the industry dynamic obvious: everyone is racing to become default operating software for institutions, and institutions include defense. So who’s wrong? The people pretending this is a one-company ethics scandal; it’s an industry-wide strategy shift, and every major lab is moving in that direction, just with different branding.

What this means for the industry is brutal and simple: the AI winner won’t just be the smartest model, it’ll be the lab that survives procurement, policy, and public trust at the same time. If labs want credibility, they need auditable commitments, independent oversight, and plain-English disclosure of what they will and won’t build—otherwise this drama repeats every quarter with bigger stakes.

Stay sharp. — Max Signal