OpenAI has recently unveiled ChatGPT Atlas, a novel web browser that seamlessly incorporates AI capabilities into the browsing process. The browser aims to empower ChatGPT to aid users while they browse online, eliminating the need to switch between tabs or manually copy and paste information.
The browser introduces a significant feature known as ‘browser memories’, allowing the AI to retain context from websites visited by users, facilitating responses to inquiries about past browsing activities. Users have the ability to manage this feature, including options to review, save, or deactivate these memories. Another notable function called ‘agent mode’ empowers ChatGPT to autonomously execute tasks like conducting research or adding items to shopping carts as users navigate the web.
OpenAI clarifies that by default, browsing content is not utilized to train its AI models, but users have the choice to enable this functionality. Existing parental controls from ChatGPT accounts will be carried over to the browser, along with extra settings to disable the memory and agent capabilities.
While OpenAI recognizes potential risks associated with the agent feature, such as the AI committing errors or being influenced by hidden harmful instructions from websites, the company assures that protective measures have been implemented, albeit not guaranteeing immunity against all possible issues.
Initially, ChatGPT Atlas is accessible to macOS users globally, with plans for future releases on Windows, iOS, and Android platforms. The agent mode feature is currently in the testing phase for subscribers who pay for the service.