OpenAI smells blood in the water

Anthropic refused to cave to Pentagon pressure demanding unrestricted military use of its AI models. CEO Dario Amodei drew two red lines: no mass domestic surveillance, and no fully autonomous weapons. Current AI systems, he argued, simply aren't reliable enough for the latter, and the former is incompatible with democratic values.

The Trump administration responded with fury. Deputy Defense Secretary Emil Michael called Amodei a liar with a God complex. The government threatened not just to cancel Anthropic's contracts, but to designate it a "supply chain risk" — a label previously reserved for Russian and Chinese companies like Kaspersky and Huawei.

As I told De Volkskrant, the real story is how the balance of power has shifted. The counterweight to government overreach is now a commercial company. That's fundamentally undesirable. Anthropic's principled stance also serves a commercial purpose — positioning yourself as the one company that won't go along with the militarization of technology is attractive branding.

But there are two separate arms races running simultaneously here that got tangled up in one week: a traditional state-level arms race over military AI, and a commercial one over model development. In the commercial race, Anthropic has already had to give ground — the costs of AI development are so extreme that they simply can't afford to hit the brakes.

Original article available here