Simulated adolescent conversations are not a clinical outcome trial. Product versions and safeguards change. The assessment also described improvements in explicit crisis handling, which should not be omitted.
What the assessment contributes
Common Sense Media’s November 2025 release describes work with Stanford Medicine’s Brainstorm Lab assessing major chatbots in simulated adolescent mental-health conversations. It reported missed distress cues and weaker behavior in extended exchanges, while also noting improvements on some explicit self-harm prompts.
Later provider announcements, including Meta’s 2026 description of parent alerts, show that safeguards continue to change. An announced feature is not independent proof of effectiveness. Neither type of evidence establishes which product is clinically safe for a particular person.
Keep healthcare operations separate from clinical authority
A practice may have useful administrative work for AI without making it a diagnostic or therapeutic tool. Public-source staff materials, nonclinical procedures, and synthetic workflow prototypes are different from advising an individual patient.
- Keep diagnoses, treatment, and crisis response with qualified professionals.
- Do not submit patient information without the required agreements and safeguards.
- Evaluate specific workflows rather than advertising an assistant as universally safe for health use.
A PRACTICAL THADDEUS WORKFLOW
Support the organization without impersonating the clinician.
Use Thaddeus for bounded healthcare operations such as organizing public guidance, drafting nonclinical training, and building administrative prototypes with fictional records. A plan can make the exclusion of clinical decisions explicit.
Try this assignment
Create a nonclinical front-desk training guide from these approved public materials. Use fictional examples, flag policy questions for our clinical lead, and exclude diagnosis, treatment recommendations, and patient records.
Thaddeus is not being presented as a clinician, therapist, crisis service, or clinically safer companion. Protected health information requires suitable written agreements, including a BAA where applicable, and confirmed technical safeguards.
Sources & further reading
- Common Sense Media finds major AI chatbots unsafe for teen mental-health support ↗
Common Sense Media. November 20, 2025 press release about the assessment. - AI risk assessment: mental-health support ↗
Common Sense Media. November 2025 assessment. - Keeping parents informed about teens’ distress conversations with Meta AI ↗
Meta. July 2026 provider announcement.
Educational information, not a security audit, professional advice, or guarantee. The assignment is an example to evaluate, not a completed demonstration. Historical incidents and product capabilities do not establish comparative safety. Confirm current policies and agreements for your intended workflow.