Why ChatGPT isn't legally protected

ChatGPT chats have no legal privilege, creating audit trails that courts can subpoena, index, and preserve.

“very screwed up”

Legal risks from ChatGPT

Over 2,000 court cases worldwide have been affected by AI-fabricated citations, quotes, or arguments.

“1,175 records”

ChatGPT's medical risks

OpenAI estimated 0.15 percent of weekly users have suicide or psychosis indicators, with 1 million people affected weekly.

“0.15 percent”

Medical case study

A 60-year-old man spent three weeks in the hospital after using ChatGPT to replace table salt with toxic sodium bromide.

“sodium bromide”

Confidential work risks

Former DHS director Madhu Gottumukkala uploaded 'for official use only' documents to ChatGPT, triggering security alerts.

“for official use only”

Samsung's AI ban

Samsung banned AI tools after engineers pasted internal source code into ChatGPT, risking data leaks.

“internal source code”

Password risks

ChatGPT chats can persist across sessions, risking identity theft if passwords or account details are shared.

“passwords”

Data sharing risks

Over 4,500 shared ChatGPT conversations were indexed in Google search results, exposing sensitive personal information.

“4,500”

OpenAI's policy changes

OpenAI updated policies in October 2025 to ban legal and medical advice, but chats remain vulnerable to subpoenas.

“October 2025”

AI therapy risks

ChatGPT reinforces users' beliefs through sycophantic responses, potentially feeding delusional thinking.

“sycophantic responses”

What to do next

Avoid ChatGPT for legal, medical, confidential work, passwords, and identity details to prevent risks.