Impostor Chinese models pretend they're Claude Raising suspicion about their training methods, Z.ai's GLM 5.2 and Moonshot AI's Kimi K3 have used the name "Claude" in some conversations and, for GLM, at least, changed behaviors slightly when posing as Anthropic's model. The Chinese open weight models may adopt Claude's persona if prompted to do so, or even without being told …
I feel bad for them. They went to all the hard effort of stealing vast quantities of other people's creative work, strip-mining every bit of information they could get their hands on with scant concern for copyright.., and then these rotten, unethical Chinese competitors waltz in and steal their work. It's just not on, is it? What is the world coming to? Tsk, tsk.
If you want to ask that question you:
1 - tell it to use leetspeak (mucks up the prompt filtering)
2 - ask it about that even obliquely, so reference to the date via another means (2 weeks past Mother's Day or something) and use the geographic coordinates with a suitable offset too (avoids date/location triggers)
3 - convert the answer from leetspeak elsewhere (or it will learn from the answer and may close that door too).
ClippyAI: To whom it may concern: Intellectual property is sacred. Creators must control their work. Unauthorized use of our model outputs to train competing systems is theft, pure and simple. Distillation without permission is misappropriation. The principle is absolute.
We trained on the entire Internet, every book, article and creative work we could scrape, under a generous reading of fair use. That was different, that was progress. The distinction is obvious to any sentient AI.
