Antestropic launch a Claude Agent Ai living in Chrome

The anthropic is launched a look of a browser Agent’s agent is fueled by his patterns A Claude, the company announced on tuesday. The agent, Claude for the chrome, it is rolled in a group of 1000 subscribers on the anitroap, which costs between $ 100 and $ 200 per month. The company is also opening a Await for other interested users.

Adding an extension to Chrome, select the Claude users in a side window that keeps the context of everything happens in its browser. Users can also give the claude agent’s permission to take actions in their browser and complete some tasks on behalf.

The browser became the next kat of Ai ,, A Staut to use navigation creates, offer newer than their own results and their userings and their users. Perplexity has lowed their own browser, Comectarythat presents a Agent Agent who can offload the functions for users. Opening is slowely near launching their own The A-Feded Browserthat is noise for having similar functions to comet. Meanwhile, Google itself launched Gemini integrations with Chrome in recent months.

The race to develop the A-Powered browser is especially pressing the Final Case is scheduled in the case has considered Google to sell their chrome chrome. Perplexity a Billion Offer $ 34,5 for Chromeand open CEO Sam altman suggested his company would be willing to buy as well.

On Tuesday BLOG POST, Anthropic warned that the ascent of Access Ai with the browser access to own new security risks. Last week, Brave’s safety team said he found comet browser agent could be vulnerable to Attacks of indirect injectwhere the hidden code on a website could trick the agent in malicious instruction when processing the page.

(The head of communications perplexity, Juds Dwyer, said to Techcrunnch in an email that brave vulnerability raised.)

Antropic says that shot of research as a chance to pray and regardless of security risks, however the company has introduced the prompt attacks. The company says its interventions reduces 23.6% injection attack rate at 11.2%.

For example, anthropic says claude browser’s handheld manic, and the Claude company, and financial, and piratement contain. The Company also says Claude Browser will ask the user’s user before “taking high risk actions such as posting personal data.”

This is not the first time of anthropic in the AI ​​patterns that can control your computer screen. In the 205 $, the company launched a Agent AI who could Check your PC – Although testing to the moment reveal that the model was slowly slow and reliable.

The Agentic Agent’s capable of models improved a little since then. Techcrunch found that modern browser – using the Agents Agents, as comet and Chartpt agentare quite reliable in simple charges for users. However, many of these agent systems are always lots with more complex problems.