Anthropic Claims Chinese AI Firms Used Claude to Train Rival Models

aeiou.pt

Factuality

Ownership

To be true, Beijing's main concern would not be the theft of intellectual property or copying of models, but data security, says an analyst. Will the Chinese companies of AI secretly travel to American models to shorten the technological distance? Is this the central question that hangs over the Chinese technological sector, after a wave of accusations from Washington and Antropic, Claude's owner. Further: when you ask a question to a Chinese AI…

DiarioBitcoin

Factuality

Ownership

Anthropic claims that Chinese developers used fraudulent accounts to access Claude and extract capabilities from the model. The company reported more than 16 million exchanges through some 24,000 accounts; separately, US agencies warned about model distillation campaigns. The accusations do not amount to an independent verification of each reported transaction.