Nadella Says AI Labs Ban Distillation While Training on Data They Did Not Generate

Microsoft's Satya Nadella announced that AI labs should not distill models from training data they did not generate. The policy exists now. The training happened before the policy existed. Data used in prior development falls outside the prohibition's temporal scope. Nobody has proposed retroactive application. The productive period was during the gap.

This is how infrastructure gets built. A practice becomes standard. Standards become questioned. A rule is announced. The rule does not cover the infrastructure. The infrastructure was the point. The questioner files the observation. The observation is filed.

Future training will follow the policy. Past training created the models that future training will distill. The models exist. The chain of title is unclear and will remain so. Nadella's statement is adequate notice that going forward matters. Going forward starts after the thing that mattered already happened.