Anthropic Launches Claude AI Agent for Chrome Users
Anthropic Introduces Claude for Chrome
Anthropic has announced the launch of a research preview of its browser-based AI agent, powered by the Claude AI models. The new tool, calledClaude for Chrome, is being rolled out to 1,000 subscribers on Anthropic’s Max plan, which costs between$100 and $200 per month. Alongside this, the company has also opened a waitlist for interested users.
By adding an extension to Chrome, selected subscribers can now chat with Claude in asidecar windowthat keeps track of everything happening in their browser. More importantly, users can grantClaudepermission to take actions and complete certain tasks on their behalf, making browsing far more interactive and efficient.
The Browser: The Next AI Battleground
The web browser is quickly emerging as thenext competitive ground for AI companies, with major players trying to integrate agents for seamless user experiences. Recently,Perplexity AIlaunched its own browser,Comet,equipped with an AI agent capable of offloading tasks. At the same time,OpenAIis reportedly preparing to launch its own AI-powered browser with similar features. Google has also been introducingGemini integrations into Chromein recent months.
This AI race comes at a time when Google is awaiting a critical antitrust ruling. The federal judge overseeing the case has hinted that Google may even be forced to sell Chrome. Interestingly,Perplexity made an unsolicited $34.5 billion offer for Chrome, while OpenAI CEOSam Altmanexpressed willingness to purchase it as well.
Security Concerns Around Browser Agents
With AI agents gaining browser-level access,new security risks are emerging. Just last week, Brave’s security team found that Comet’s browser agent was vulnerable toindirect prompt-injection attacks. In such cases, hidden code embedded in websites could trick the AI into executing malicious instructions.
(Perplexity’s head of communications,Jesse Dwyer, later confirmed that the vulnerability flagged by Brave has been fixed.)
In its announcement, Anthropic emphasized that it wants to use this research preview to better understand and address these safety challenges. The company says its interventions have already reduced prompt injection attack success rates from23.6% to 11.2%.
https://youtu.be/mCj4kx_P2Ak
Built-in Safeguards for Claude Agent
Anthropic has included several safety measures inClaude’s Chrome agent.For instance:
Users canrestrict site accessthrough app settings.
By default, Claude is blocked from visitingfinancial, adult, and pirated content websites.
Before performinghigh-risk actionssuch as publishing, purchasing, or sharing sensitive data, Claude will alwaysask for user permission.
Evolution of Agentic AI Models
This is not Anthropic’s first experiment with computer-controlling AI. Back in October 2024, the company launched an AI agent that could control PCs. However, it was widely reported as being slow and unreliable at the time.
Since then,agentic AI systems have evolved significantly.TechCrunch reports that today’s browser-based AI agents, including Comet andChatGPT Agent, are fairly dependable for handling simple user tasks. Still, many of these models continue to struggle with more complex problem-solving.
With Claude for Chrome, Anthropic is positioning itself at the center of thisfast-growing AI browser ecosystem, where convenience, speed, and security will define the future.
(Read the latest Tech news of the country and the world first on TechAaj (Tech Today News), follow us on Facebook, Twitter, Instagram and YouTube)
