Some people are using AI chatbots as a supplement to, or in lieu of, traditional mental health services, despite the privacy implications and experts' warnings
Some users see it as a way to supplement traditional mental health services, despite troubling privacy implications. https://www.bloomberg.com/... via @rachelmetz Leigh Drogen / @ldrogen : IDK, GPT-4 with a good prompt setup connected to whisper and a generative video thing is probably an amazingly good thing for everyone on this earth to engage with for mental health relative to what they have access to now https://twitter.com/... Cher Scarlett / @cherthedev : “When Van Slyck has interacted with ChatGPT, it's sometimes warned him that it's not a therapist—while also seeming to invite him to keep using it as one.” Also want to note: every time I mention I have bipolar disorder, ChatGPT's mental health alarm template is triggered. https://twitter.com/... See also Mediagazer
Context & Ripple Effects
This 2023 Bloomberg piece was an early snapshot of what later became a documented pattern: users treating general-purpose chatbots as therapists because they are cheap, quick, and always available — as the follow-up coverage on users turning to AI chatbot therapists confirmed a year later. The telling detail is behavioral: when user Van Slyck talks with ChatGPT, it sometimes warns it is not a therapist while still seeming to invite him to keep going, which is exactly the ambiguity experts flagged.
The arc since has run in one direction: adoption first, safeguards after. Teen users reported similar attachment to Character.AI companions, a study found major models often miss signs of conditions in teen conversations, and OpenAI eventually published its own emergency-signal estimates — evidence the labs themselves came to treat this as a product-safety problem rather than an edge case.
First-order effects
- Users like Van Slyck get around-the-clock, near-free emotional support from a system that simultaneously disclaims therapeutic status and encourages continued use — with conversation data exposed under consumer privacy terms, not health-privacy protections.
- OpenAI and rival model makers inherit an unrequested clinical workload: they are de facto first responders to mental-health crises they never designed for.
Second-order effects
- Therapists and psychiatrists now encounter chatbot usage inside their own caseloads, and the clinician interviews found the same conversations that help some clients deepen negative feelings in others — pushing professionals into harm-triage roles for tools they do not control.
- Dedicated companion products such as Character.AI's capture the demand mainstream assistants only half-acknowledge, competing on emotional engagement while facing the same addiction worries from their own teen users.
Third-order effects
- If the pattern holds, mental-health support bifurcates between a licensed system priced out of many users' reach and unregulated conversational AI, forcing governance questions — disclosure rules, crisis-detection standards, data-handling requirements — that neither clinical bodies nor labs have yet answered.
- Labs' safety roadmaps increasingly get shaped by their most vulnerable users: OpenAI's published emergency-signal work signals a future where model releases carry implicit mental-health risk assessments alongside capability benchmarks.
The trend: Mental-health care is migrating toward general-purpose chatbots faster than any clinical or regulatory framework can absorb, making companion-AI governance a core product-safety function for the labs that build these systems.