The San Francisco-based startup also said it planned a "ChatGPT Business" subscription with additional data controls.

The move comes as scrutiny has grown over how ChatGPT and other chatbots it inspired manage hundreds of millions of users’ data, commonly used to improve, or "train", AI.

Italy last month banned ChatGPT for possible privacy violations, saying OpenAI could resume the service if it met demands such as giving consumers tools to object to the processing of their data. France and Spain also began probing the service.

Mira Murati, OpenAI's chief technology officer, told Reuters the company was compliant with European privacy law and is working to assure regulators.

The new features did not arise from Italy's ChatGPT ban, she said, but from a months-long effort to put users "in the driver's seat" regarding data collection.

"We'll be moving more and more in this direction of prioritizing user privacy," Murati said, with the goal that "it’s completely eyes off and the models are super aligned: they do the things that you want to do".

User information has helped OpenAI make its software more reliable and reduce political bias, among other issues, she said, but added that the company still has challenges to tackle.

Tuesday's product release lets users switch off "Chat History & Training" in their settings and export their data.

Nicholas Turley, the OpenAI product officer who likened this to an internet browser's incognito mode, said the company still would retain conversations for 30 days to monitor for abuse before permanently deleting them.

In addition, the company's business subscription available in the coming months will not use conversations for AI model training by default.

Microsoft Corp (MSFT.O), which has invested in OpenAI, already offers ChatGPT to businesses. Murati said that service would appeal to the cloud provider's existing customers.

Reporting By Jeffrey Dastin in Palo Alto, Calif. and Anna Tong in San Francisco; Editing by Sonali Paul (REUTERS)