Anthropic is exploring an interesting road in making its AI assistant more practical by bringing it directly into Chrome. The company today announced Claude for Chrome, a browser extension that allows the AI to see what users are looking at, click buttons, and even fill forms on their behalf.
We’ve developed Claude for Chrome, where Claude works directly in your browser and takes actions on your behalf.
— Anthropic (@AnthropicAI) August 26, 2025
We’re releasing it at first as a research preview to 1,000 users, so we can gather real-world insights on how it’s used. pic.twitter.com/lVDKhnPbHY
Currently, the tool is launching as a research preview for 1,000 Claude Pro (Max plan) users, with a waitlist open for others. Anthropic says the limited rollout is meant to gather real-world feedback on how people use the feature, what shortcomings appear, and most importantly, how safe it really is.
On paper, this looks like a smart move, especially when compared to projects like Comet, which require users to install an entirely new browser with a different interface. In contrast, Claude for Chrome integrates into the browser we are already using, making it far less disruptive. That said, I haven’t tested Claude for Chrome myself, so I can’t vouch for how well it works in reality.
The bigger challenge here is safety. Browser-using AIs face risks such as prompt injection attacks, where malicious actors hide instructions inside websites, emails, or documents to trick the AI into taking harmful actions. Anthropic says internal “red-teaming” tests showed that without protections, Claude could be fooled into deleting emails, stealing data, or even making financial transactions.
According to the shared press release, in 123 test cases across 29 attack scenarios, browser use without safeguards had a 23.6% attack success rate. With safety measures in place, that number dropped to 11.2%, and for some browser-specific attacks, the rate went from 35.7% to 0%.
On top of that, Anthropic has built some defenses into Claude for Chrome, which include Site-level permissions, Action confirmations, Restricted categories, and Suspicious activity detection. However, the company admits that the defenses aren’t perfect yet, which is why it’s starting with trusted users in controlled conditions.
Interested users can join the waitlist at claude.ai/Chrome. Once accepted, they’ll be able to install the extension from the Chrome Web Store and log in with their Claude credentials.
