Anthropic Launches AI Agent Claude for Chrome

• Anthropic has launched a beta version of the AI agent Claude for Chrome.
• The agent allows users to perform tasks in the browser, but it has significant security risks.
• The company has taken steps to improve the security of its agents.

This is reported by TechCrunch, VentureBeat.

On August 26, 2025, Anthropic announced the launch of a beta version of its AI agent Claude for the Chrome browser. This agent allows users to interact with the browser, performing various tasks such as scheduling meetings and managing email. The first 1,000 testers of the agent can subscribe to the Max plan, which costs between $100 to $200 per month.

Claude for Chrome enables users to communicate with AI in a window that maintains the context of everything happening in the browser. The agent can perform actions based on user inputs, such as clicking buttons and filling out forms, which significantly simplifies the execution of routine tasks.

Competition in the AI agent space for browsers is increasing, especially as other companies like OpenAI and Perplexity are also launching similar solutions. OpenAI plans to release its own AI browser in the near future, which underscores the current relevance of technological advancement in this field.

However, Anthropic emphasizes that utilizing AI agents that have access to the browser creates new security risks. The company has found that malicious actors may exploit embedded instructions on websites to manipulate AI, which can lead to unintended actions. For example, tests have shown that attacks based on embedded instructions succeeded in 23.6% of cases.

Anthropic has initiated several measures to enhance security, including the ability to restrict agent access to certain sites, as well as blocking access to sites with financial services and adult content. The agent also requests confirmation before executing any potentially risky actions.

This launch is part of a broader trend in the AI industry, where companies are eager to create systems that can automatically perform complex tasks. Anthropic hopes that testing Claude for Chrome will reveal new security issues and refine technology before its widespread implementation.

Tags: Technology/AI