Claude for Chrome is here and can be used directly as a browser extension.

On the 27th, Anthropic released its latest browser agent - Claude for Chrome!

This agent tightly integrated with Chrome can be directly used as a browser extension:

It can not only help you set up your calendar.

It can also help you reply to emails.

It can even handle house hunting with ease!

This left everyone stunned, so much so that a netizen directly posed a soul - searching question:

Since AI functions can be achieved through browser extensions, is there still a need to open a separate browser?

So...

Perplexity: Oh no, after waking up, am I the clown?

(Note: Perplexity recently launched its own browser, Comet, in the bottom - right corner of the above picture)

Claude for Chrome: Safety First

As an agent based on Chrome, Claude for Chrome allows users to add extensions to Chrome.

We can have a conversation with it in the side window, and this chat box will retain the context of all activities in the browser.

In addition, users can authorize the agent to perform operations in the browser and let it complete certain tasks on their behalf, as shown at the beginning.

Currently, this mini - program is only open to 1000 selected Max package users. Other users can subscribe to the Max package and join the waiting list. According to the Claude official website, the monthly fee for this package ranges from $100 to $200.

As for why the test scope is so limited, Claude says: Safety comes first!

We believe that using artificial intelligence in browsers is inevitable: a large amount of work is done in browsers, so enabling Claude to see what you're viewing, click buttons, and fill out forms will make it even more useful. However, using artificial intelligence in browsers poses security challenges that require stronger safeguards.

In other words, Claude collects data and security feedback through a small - scale pilot, optimizes classifiers and models, and upgrades browser security and functionality simultaneously.

Specifically, the most troublesome vulnerability for AI browsers at present is the so - called "prompt injection attacks".

These attacks hide instructions in websites, emails, or documents, tricking AI into performing harmful operations such as deleting files, stealing data, or conducting financial transactions without the user's knowledge.

Before the release of Chrome for Claude, a malicious email once made Claude delete an email without the user's permission.

In response, Anthropic has conducted extensive tests and continuously formulated security measures to meet users' needs for safe Internet access.

For example, Anthropic says that users can restrict Claude's browser agent from accessing specific websites in the app settings.

Meanwhile, Claude also prohibits access to websites that provide financial services, adult content, and pirated content by default.

Finally, Anthropic emphasizes that Claude's browser agent will first obtain the user's permission before performing "high - risk operations such as posting information, shopping, or sharing personal data".

It's worth mentioning that this isn't Anthropic's first attempt to develop an agent that can control the computer screen.

As early as October 2024, the company launched an agent that could control a PC, but it was shelved due to speed and stability issues.

Now, this kind of exploration is no longer exclusive to Anthropic. AI browsers are gradually becoming the next battlefield for AI giants.

The Browser: The Next Battlefield for AI Giants

As Anthropic said, when artificial intelligence can interact with web pages, it will create extraordinary value.

As a bridge between users and information, AI giants hope to make the connection between AI and users closer through browser integration.

For example, Perplexity recently launched its own AI browser, Comet. Similar to Chrome for Claude, it can perform functions such as booking meetings, following up on emails, browsing other pages, conducting research, shopping, and summarizing.

Meanwhile, Google and Microsoft will embed Gemini and Copilot in their Chrome and Edge browsers respectively.

In addition, it is reported that OpenAI is also about to release an AI - driven browser with functions similar to Comet.

However, at this point, there seem to be several different paths:

One is to, like Anthropic, develop an extension and directly integrate the agent into an existing browser.

The other is to, like Perplexity, start from scratch and build a new browser and then embed its own AI into it.

For giants like Google and Microsoft, which have both AI and browsers, there are even more options.

Of course, all of this is based on the premise that your browser has users - which is also one of the reasons why the future of Chrome has attracted so much attention.

(Note: In 2025, Google Chrome accounted for 68.35% of the global browser market, far exceeding Safari in second place, which had 16.25%)

So, where exactly are AI browsers headed?

Reference Links

[1]https://x.com/AnthropicAI/status/1960417002469908903

[2]https://claude.ai/chrome

[3]https://www.anthropic.com/news/claude-for-chrome

[4]https://techcrunch.com/2025/08/26/anthropic-launches-a-claude-ai-agent-that-lives-in-chrome/

This article is from the WeChat official account "Quantum Bit", author: henry, published by 36Kr with authorization.