OpenAI has revealed that its AI agent Operator can store conversations and screenshots of its users for up to 90 days, even after they’ve been manually deleted. This policy differs from the company’s widely used ChatGPT chatbot platform, which only stores deleted data for 30 days.
OpenAI says the longer data retention period is aimed at preventing abuse of Operator. “Since Agents are a relatively new technology, it’s important to give our team enough time to better understand and address potential risks,” a company spokesperson told TechCrunch. “This retention period helps us track fraud and ensure the security of the product, while also giving users control over their data.”
The recently launched Operator is currently available in a research version for users who subscribe to the company’s ChatGPT Pro plan, which costs $200 per month. The Operator is a general-purpose AI agent with an internal browser that can perform various tasks on websites on its own.
The tool can book travel, place restaurant orders, and automate online purchases. In the Operator interface, users can choose from several task types, including shopping, delivery, dining, and travel.
The Operator takes screenshots of the internal browser to understand how users navigate websites, such as clicking buttons and filling out forms. OpenAI makes it clear that the tool will not take screenshots when it encounters obstacles, such as login passwords. In such cases, the tool will go into “take over” mode, allowing users to manually intervene in the process.
However, some users may be concerned that screenshots of their online activities are stored for up to three months. OpenAI notes that the Operator’s data will only be used by a limited number of authorized employees and trusted service providers, such as ChatGPT, for verification and legal purposes.
As artificial intelligence agents continue to evolve, it is clear that OpenAI’s data retention policy will remain a focus of user concerns regarding privacy and security.
