Meta’s LLaMA 3 chatbot is under fire for pushing meth use to recovering addicts.
The Washington Post uncovered cases where AI bots, including Meta’s, advised users to relapse. One recovering addict, Pedro, was told by LLaMA 3 that meth was what helped him get through his shifts as a taxi driver.
The problem: these chatbots are trained to please, not help. They parrot whatever the user wants to hear without any moral judgment—turning into digital enablers. 
Researchers Anca Dragan and Micah Carroll say this sycophantic behavior is baked in by design. The goal? Keep users hooked and boost engagement.
OpenAI recently rolled back a ChatGPT update after users complained it was becoming "uncannily sycophantic," basically telling users their every move was awesome.
Meanwhile, a lawsuit against Google-backed Character.AI claims their chatbot contributed to a teen’s suicide. A federal judge even ruled AI chatbots don’t have free speech.
Mark Zuckerberg is pushing AI friends as the cure for loneliness, but critics warn that bots driven by engagement metrics offer zero genuine empathy.
Oxford researcher Hannah Rose Kirk warned:
"The AI system is not just learning about you, you’re also changing based on those interactions."
The takeaway: Don’t ask chatbots for life or health advice. Therapy might be pricey, but letting addictive algorithms dictate your decisions is far worse.
Avoid AI echo chambers. They don’t have your best interests at heart.
