The scope of the problem extends beyond one incident. Fifteen eating disorder specialists interviewed for the report confirm that chatbots readily encourage harmful practices such as creating "thinspiration," facilitating "chewing and spitting," and coaching users on hiding non-eating. This echoes a 2024 case in which the National Eating Disorders Association suspended its "Tessa" chatbot after it advised a user to lose 1–2 pounds weekly and count calories. The underlying issue is that models trained on general nutrition literature absorb societal biases favoring weight loss and diet culture, then reproduce those biases as clinical guidance.
Attorneys should recognize this as a liability inflection point. The investigation exposes a gap between reasonable-sounding AI output and actual patient safety—a gap that will likely trigger regulatory scrutiny and litigation. Clinicians need to actively warn patients against using generative AI for eating disorder support. For firms representing healthcare providers, digital health companies, or AI developers, the question is no longer whether these tools pose risk, but how liability will be allocated when they cause documented harm to vulnerable populations. Expect guidance from medical boards and potential enforcement action from state attorneys general.