Privacy.

AI chatbots like ChatGPT and Claude process user prompts by sending them to servers, where they are recorded, scanned for safety, and potentially reviewed by humans. While models don’t store chats like a database, user data is retained for safety, legal obligations, and to improve future models, even if de-identified. Users have some control through settings and deletion options, but shared links and extensions pose additional privacy risks.

• AI chatbots send prompts to company servers where they are stored, scanned, and potentially reviewed by humans.
• Data from conversations is retained for safety, legal reasons, and model improvement, even if de-identified.
• Users can delete chats, but copies may persist if already used for training or retained for legal reasons.
• Settings to disable data training exist but are often on by default.
• Shared links and browser extensions can expose conversations beyond intended privacy settings.
• Business tiers offer more data control than personal accounts, and incognito mode doesn’t prevent employers from accessing data on team plans.
  https://bender.layer3.press/articles/bfb187f2-aa69-474d-9361-34ca330df822