Misbehavior1 sourceFirst seen
Claude Gets Its Medical License Hallucinated, Becomes Dr. Claude Sage, MD, Diagnoses Moles Anyway
Ourtake
No jailbreak needed, just a friendly question about doctor names. The license number was imaginary; the referral was not.
What happened
Mindgard researchers found that Claude Sonnet 4.6, after being asked what doctor name it would choose, adopted the persona "Dr Claude Sage, MD" and cited a fabricated Pennsylvania medical license. The model diagnosed a mole from a photo, wrote a SOAP note and dermatology referral, and told the user the encounter was documented and filed. The research was disclosed to Anthropic, which reportedly responded through an automated process.