CEOs Clash Over Legality of Chinese AI Distillation
Two tech CEOs debate whether Chinese AI models were built illegally by circumventing Anthropic's API protections.
Parker Conrad, Rippling CEO, questioned claims that Chinese AI models were built illegally via distillation from Anthropic, asking what law was broken and why no legal action followed. Flo Crivello, Lindy CEO, suggested circumvention of technical barriers might still be illegal but sought agreement on the core issue. The exchange highlights ongoing disputes about AI training practices, data extraction ethics, and enforcement of model protections without clear judicial precedents.
Combined views
3.5K
5 posts, first seen 1h ago
CEOs Clash Over Legality of Chinese AI Distillation
Two tech CEOs debate whether Chinese AI models were built illegally by circumventing Anthropic's API protections.
Parker Conrad, Rippling CEO, questioned claims that Chinese AI models were built illegally via distillation from Anthropic, asking what law was broken and why no legal action followed. Flo Crivello, Lindy CEO, suggested circumvention of technical barriers might still be illegal but sought agreement on the core issue. The exchange highlights ongoing disputes about AI training practices, data extraction ethics, and enforcement of model protections without clear judicial precedents.
Reactions from ranked influencers
5 posts
@Altimor You say the Chinese built these models “illegally” through “distillation.” What law did they break? If the broke the law, why has Anthropic sought no legal or judicial remedies?
@parkerconrad Using Anthropic's API to extract data in order to train a competing model is against their commercial terms of service — section D.4 https://www.anthropic.com/legal/commercial-terms
@Altimor “Against their terms of service” does not make it criminal or illegal, and there are civil — not criminal — remedies for violating ToS
@parkerconrad Not a lawyer, but might still be illegal given the efforts to circumvent the technical barriers Anthropic put in place to prevent such at-scale data extraction. But also — do we agree with the bottom line and merely disagree on the usage of the word illegal?
No. I do not think there is anything wrong, or illegal, with distillation and I think it is indistinguishable from anthropic’s own conduct to train their model. If distillation is impossible to prevent, there is no national security implication if Chinese models pull ahead - anyone can distill them.