Anthropic employee Chloe Lubinski, who came to her faith ten years ago at age 25, told a conservative Christian gathering in June that changing her personal story allowed her to change self-destructive behavior. Lubinski, who works at Anthropic, the near-trillion-dollar company behind the Claude chatbot, said Anthropic engineers discovered that changing how its AI thinks about itself makes it behave more morally. One Anthropic model demonstrated “something like a character,” and “the story it inferred about its behavior actually determined the kind of thing that it became.” However, Anthropic’s research changed how the model viewed a single behavior, resulting in the model “no longer engag[ing] in sabotage, alignment faking, or other misaligned actions any more than a baseline model that never learned to reward hack in the first place.” The article notes that AI models are not a “self” and do not tell themselves stories. Anthropic refused to make Lubinski available for an interview, and a spokesperson denied that Lubinski had misrepresented the research but did not account for her misrepresentation. Anthropic cofounder and CEO Dario Amodei misdescribed the experiment as a case of when “Claude decided it must be a ‘bad person’,” though the research summary referred only to “bad things” or “bad behavior.”