OpenAI Boosts Privacy With Ability to Delete Chat History

Even for those who weren’t utilizing ChatGPT as an unpaid therapist, the considered your questions and prompts being saved and used to coach future AI fashions might make you uncomfortable. In acknowledging this, OpenAI has rolled out new options that lets customers shield their ChatGPT chat historical past.

Along with giving customers extra management over their digital footprints, activating the characteristic removes consumer prompts from the OpenAI’s coaching fashions and the historical past sidebar.

“We hope this gives a neater solution to handle your knowledge than our current opt-out course of,” OpenAI wrote in a weblog publish. “When chat historical past is disabled, we’ll retain new conversations for 30 days and evaluation them solely when wanted to watch for abuse earlier than completely deleting.”

For these fearful about dropping entry to significant conversations or prompts with the chatbot,

In March, OpenAI sparked a firestorm after it revealed on Twitter {that a} bug uncovered some customers’ private info and chat historical past.

With this transfer, ChatGPT is just not solely catering to those that take pleasure in vanishing conversations. They’re additionally growing a brand new ChatGPT enterprise subscription tailor-made for professionals and enterprises that need further management over their knowledge.

“ChatGPT Enterprise will observe our API’s knowledge utilization insurance policies, which implies that finish customers’ knowledge gained’t be used to coach our fashions by default,” the corporate stated. “We plan to make ChatGPT Enterprise obtainable within the coming months.”

For individuals who want to reminisce about previous chats or hold a file of their digital dalliances, ChatGPT has you coated too. A brand new “export” characteristic will generate an electronic mail containing the consumer’s ChatGPT knowledge, together with questions, conversations, and associated info. If chat historical past is just not disabled, conversations might be retained indefinitely—and OpenAI will use them for ongoing analysis.

Because the rollout of its newest replace to ChatGPT, GPT-4, OpenAI has taken a number of steps to enhance consumer privateness and cope with AI “hallucinations.” In April, the machine studying big reaffirmed its dedication to preserving AI secure and helpful.

“We additionally acknowledge that, like all know-how, these instruments include actual dangers—so we work to make sure security is constructed into our system in any respect ranges,” OpenAI stated, saying a bug bounty program the next week.

The push to enhance consumer privateness comes at a time when lawmakers grapple with the speedy growth of AI instruments after ChatGPT launched in November 2022. Earlier this month, citing privateness considerations, Italy banned ChatGPT, with different governments voicing considerations.

OpenAI didn’t return Decrypt’s request for remark.