Artificial Intelligence

Mainstream outlets reported that on July 23, 2026 OpenAI expanded "Health in ChatGPT" to U.S. users 18+ across Free, Go, Plus and Pro tiers on web and iOS, letting people upload medical records and link Apple Health so the assistant can summarize and compare lab results and medication lists; OpenAI said those records would not be used to train models or target ads and that extra encryption and confirmation safeguards were added. Coverage also flagged expert warnings about risks from AI-generated medical advice (citing a Florida lawsuit and a California teen death), and noted tensions around accuracy, clinician handoffs and user trust — echoing OpenAI’s earlier January rollout and claims that hundreds of millions ask health questions on ChatGPT weekly.

What mainstream reporting generally omitted were regulatory, technical and evidentiary details that would change risk assessment: independent validation results (error rates, benchmark performance of the GPT-5.6 health variants), specifics of the encryption/safety architecture, whether HIPAA applies or how legal liability would be allocated, and independent security/privacy audits. Alternative factual sources pointed out an important legal/regulatory gap — consumer-facing LLM health guidance typically falls outside FDA medical-device oversight if presented with disclaimers — a context missing from many mainstream pieces. There were no notable contrarian commentaries surfaced in the sample, and readers would benefit from more stats and peer‑reviewed studies on harms and effectiveness, transparency on data retention/third‑party access, and independent audits before relying on summaries that focus mainly on features and company assurances.