Comment by mikesurowiec

From the full announcement

> The new classifier also comes at the cost of flagging benign requests more often during routine coding and debugging tasks. As with all our safeguards, we’ll continue to refine this to better distinguish genuine misuse from legitimate requests and reduce false positives.