OpenAI is considering encryption for ChatGPT

OpenAI is seriously weighing the addition of encryption to ChatGPT, CEO Sam Altman told reporters during a briefing, Axios reports.

The rollout is expected to start with temporary chats that aren’t saved in user history. Altman stressed the company’s commitment but noted there’s no clear timeline yet. OpenAI declined to comment further.

Altman pointed out that users share sensitive information with ChatGPT, yet these conversations don’t carry the same legal confidentiality as those with doctors or lawyers. Temporary chats seem like a logical first step for encryption, since OpenAI doesn’t use them for model training.

However, temporary and deleted chats are currently subject to a court order issued in May that requires OpenAI to retain the content of such conversations. The ruling came as part of a copyright infringement case and has drawn criticism from privacy advocates.

Implementing full end-to-end encryption for chatbots is technically complex. Unlike messengers, where the provider acts as an intermediary, in the case of AI the company is a participant in the conversation and must access the data for its algorithms to function.

Apple has partly addressed this challenge with its Private Cloud Compute technology for Apple Intelligence, which processes requests on company servers without granting broad access to user data.