Google's Project Jarvis AI just leaked an internal preview of this AI agent as a Chrome extension. 

The cynics will all call this an “intentional” leak, meant as a stealth preview. Don’t be fooled by this though, it’s too early to tell if this was an intentional preview or just an employee mistake.

Most of the time in cases it’s just an employee slip up, which is of course far less juicy of a story.

Here’s what we know so far: Google unintentionally made an internal preview of Jarvis AI available as a Chrome extension. This tool is designed to automate web-based tasks like booking flights or shopping. 

Due to lack of proper access permissions, users who downloaded it weren't able to use it.

Jarvis AI is setting its sights on assisting users by performing tasks for them  within the Google Chrome browser. It operates by capturing frequent screenshots of the user's screen to interpret and act on the content, such as clicking buttons or filling text fields.

Google was originally planning on an official release date of Jarvis in December 2024, but has supposedly been pushed out until next year. This leak comes ahead of its official introduction, possibly as part of the rollout of the next version of Google's Gemini AI model.

Anthropic’s Claude has recently been proven to already be showing AI agent capabilities.

Posts on X (formerly Twitter) have confirmed this leak, with users expressing interest and some apprehension about Google's AI capabilities, especially concerning privacy and automation.