OpenAI Launches Atlas Browser with ChatGPT Integration, Prompting Data Privacy Discussions
OpenAI has introduced Atlas, a new web browser that integrates its ChatGPT technology. The browser is currently available for Apple computers.
Browser Functionality
Atlas is designed with an "agentic mode" feature, which enables it to perform various actions on behalf of the user. Demonstrated capabilities include analyzing an online recipe, calculating required ingredients for a specified number of diners, and facilitating the online purchase of those ingredients. OpenAI CEO Sam Altman commented that artificial intelligence presents an opportunity to redefine the browser experience.
Data Privacy Considerations
The integration of ChatGPT within Atlas has led to discussions regarding user data privacy. Anil Dash, a tech entrepreneur, suggested that the extensive data requirements of large language models might be addressed by Atlas's design, which could facilitate increased access to user data. The browser is capable of interacting with user email and Google Docs, and can retain "browser memories" from visited websites. Dash indicated that the system might transmit more information to OpenAI than it provides to the user.
Lena Cohen, a Technologist at the Electronic Frontier Foundation (EFF), raised concerns about AI browsers operating in agentic mode. Cohen noted that users might transfer more control to OpenAI than initially perceived, and that managing data once it is stored on OpenAI’s servers could become complex.
OpenAI has stated that its default setting does not utilize information accessed via Atlas for training its AI models; however, users have the option to consent to this use.
Security Concerns: Prompt Injections
Experts have identified "prompt injections" as a potential security risk in AI browsers. Cohen described prompt injections as malicious instructions embedded within web pages that an AI agent could be induced to execute. Examples include an agent being directed to purchase an unintended product or to disclose credit card information. OpenAI acknowledges prompt injection as an unresolved issue and reports ongoing efforts to train its models to disregard such instructions.
Industry Perspective
Chirag Shah, a professor at the University of Washington's Information School, commented on the rapid development of AI with minimal regulatory frameworks, noting potential implications for users.