Data privacy expert warns Clark County residents on AI chatbot risks

Information typed into artificial intelligence chatbots can be stored by the companies that build them. Depending on specific service settings, that data may also serve to train new models. U.S. PIRG data privacy advocate R.J. Cross recently detailed these risks for local audiences.

Cross noted that details shared with a chatbot have appeared in unrelated online searches in recent cases. He explained that private conversations do not automatically become public, but the storage of data creates a pathway for exposure.

The privacy advocate highlighted a gap in user agreements. Many chatbots inform users that their data will be used for training during sign-up. However, they often fail to state that the data is also stored on servers where it can surface in other chats or get exposed to hackers.

"Sometimes the chatbots will say when you're signing up for account, hey, we're going to use your data for training. But it's not saying this also means we're going to store your data on our servers and that could be surfaced in other people's chats or it could be exposed to hack," Cross said.

Cross recommends avoiding sensitive inputs in these tools. Users should keep Social Security numbers, medical information, and personal documents out of the text box. He also advised reviewing privacy settings and limiting the personal details shared about friends and family.

"You don't want to give it the highly sensitive stuff that you wouldn't want ending up in the hands of someone else because using those chatbots come with those risks," Cross said.