Engadget warns against putting sensitive information into ChatGPT
The publication says chatbot conversations lack legal and medical confidentiality, and advises against sharing private details or relying on AI for professional guidance.
Engadget has warned readers against entering sensitive information into ChatGPT, citing privacy, confidentiality and reliability risks. Its advice covers legal disputes, therapy, health concerns, confidential workplace material, passwords, identity documents and financial details.
The article says conversations with a chatbot do not carry the legal or medical protections that apply when speaking with a lawyer, doctor or therapist. It notes that chats may be stored or disclosed in some circumstances, including through legal processes, and that privacy exposure can depend on product settings and events.
Engadget points to court cases in which judges found fabricated citations, quotes or arguments in filings that used AI. It also describes a case in which a man was hospitalised after following chatbot advice about replacing salt with sodium bromide. The publication’s examples were not independently verified in the supplied material.
For workplace data, the article cites past incidents involving Samsung engineers who pasted internal code into ChatGPT and a former US cyber security official who uploaded documents marked for official use. It warns that material sent to a third-party chatbot may be difficult to retrieve or delete and could be exposed.
Engadget also discusses OpenAI policy changes, including restrictions on tailored legal and medical advice, revised crisis responses and parental controls. It says a court order requiring preservation of consumer chat logs was later terminated, while noting that retained logs from the order period remained accessible to plaintiffs.
The article’s practical advice is to keep personal case details, health information, employer-controlled material and account credentials out of prompts. It cautions that the risks vary by settings and circumstances, and does not claim all ChatGPT conversations are public.
