The 'Open AI' Debate Just Split Into Two

The "open AI" debate was never one debate. It is two arguments that share a name, and this week they publicly separated.

The first is about openness as a development practice: should model weights be released at all. The second is about openness as an export question: should frontier weights cross a geopolitical border to a rival power. The two have been fused for years under the slogan "open vs. closed," which is why the public has been arguing past itself.

Dario Amodei's letter, published the same day Nvidia, Microsoft, Meta, and eventually OpenAI were signing a public defense of open-weight models, draws the line out loud. "Some people have even accused Anthropic of wanting to ban open-weights models as a means of protecting our business," Amodei wrote. That is the wrong fight, he argues; the real one is whether frontier open weights should reach China. Anthropic's conspicuous absence from the industry letter now reads as the policy line, not the PR stumble.

The leading counter is that the split is rhetoric over a familiar IP-protectionism play: Anthropic's distillation accusations against Chinese labs remain contested, so the bifurcation may be window dressing rather than a real policy line. The reader's test is the same either way: when "open AI" appears in a headline, ask which of the two debates is on the table. The slogan was a media convenience. The reality has always had two axes.