Anthropic Rolls Out Claude Sonnet 4.6 With Enhanced Computer-Use Capabilities

Anthropic Rolls Out Claude Sonnet 4.6 With Enhanced Computer-Use Capabilities

Anthropic has unveiled a new artificial intelligence model, Claude Sonnet 4.6, designed to carry out more complex computer-based tasks, marking a significant step in the company’s push to make AI tools more autonomous and task-oriented.

Set to roll out on Tuesday, Sonnet 4.6 can perform multi-step actions on a user’s computer, such as filling out web forms and coordinating information across multiple browser tabs. In a blog post, Anthropic acknowledged that the model still trails highly skilled humans in using computers but said the pace of improvement remains striking.

The company first introduced its “computer use” feature in late 2024, enabling AI systems to analyse users’ screens and execute actions such as browsing the web. Since then, rivals including OpenAI and Google, owned by Alphabet Inc., have launched similar tools aimed at automating everyday digital tasks.

OpenAI recently hired the creator of OpenClaw, an open-source tool that operates directly on a user’s computer to perform tasks such as sending emails or making reservations with the help of a connected AI model—highlighting the intensifying race in this space.

Government Issues High-Severity Alert For Google Chrome Users

Claude Sonnet 4.6 is now the default model for users of Anthropic’s Claude chatbot, both on the free tier and the Pro subscription plan. The company also said the new version delivers improved reliability in coding, an area where Anthropic has long focused its development efforts.

Market Impact and Security Concerns

Anthropic’s broader expansion beyond software development has unsettled investors in recent weeks. A quiet release of a tool aimed at automating certain legal tasks contributed to sharp declines in shares of some software firms, amid fears that AI-driven automation could disrupt traditional service providers.

Financial services stocks also came under pressure after Anthropic introduced an updated version of its Opus model, tailored to financial research. The market reaction underscores growing anxiety over which sectors may be most vulnerable to rapid AI-driven transformation.

OpenClaw Rolls Out Major Updates, Makes Multiple AI Models Free and Easier to Access

At the same time, expanding AI models’ ability to act independently on computers introduces new risks. Delegating control to such systems can expose users to security threats, including prompt injection attacks, where malicious instructions manipulate an AI model’s behaviour. Anthropic said Sonnet 4.6 demonstrates stronger resistance to such threats compared to its predecessor, Sonnet 4.5.

As AI systems take on increasingly sophisticated digital tasks, the balance between productivity gains and security safeguards is likely to remain central to the debate.