I wish Anthropic or someone would take a leadership role and re-train their models without any GPL code, or at least stop doing so in the future tense.
Opensource has provided training data, closed source code not so much. They will have a bad time learning.