**Google Unintentionally Reveals AI Agent Capable of Browsing the Internet**
Google has unintentionally disclosed its forthcoming AI agent, referred to internally as “Jarvis,” which is intended to navigate the web on behalf of users. As reported by *The Information*, a “preliminary preview” of this AI agent was briefly accessible for download as a Chrome extension on Tuesday. The extension was characterized as “a supportive companion that explores the web alongside you.”
Even though the prototype was available for download, it was non-functional due to a lack of necessary access permissions. The extension was subsequently taken down from the Chrome Web Store that same day.
**SEE ALSO:**
[Google Maps is utilizing Gemini to assist you in planning](https://mashable.com/article/google-maps-gemini-new-features)
This unintentional exposure supports previous claims regarding Google’s ongoing developments in AI agents. Jarvis is anticipated to aid users by browsing the internet, making purchases, and securing flight bookings. Google is reportedly set to officially launch Jarvis in December, coinciding with the release of its newest version of the Gemini large language model.
Google is not the only technology firm working on AI agents. Last month, Anthropic unveiled a feature for its Claude LLM that can effectively [control a computer](https://mashable.com/article/anthtopic-claude-computer-use-ai) to read and write JavaScript code. OpenAI is also engaged in the creation of comparable AI agents. In a recent [Reddit AMA](https://mashable.com/article/openai-sam-altman-ama-reddit), OpenAI’s Chief Product Officer Kevin Weil suggested that ChatGPT may soon have the ability to initiate conversations with users, with task automation being a “major focus in 2025.”