Bỏ qua đến nội dung chính
Back to home
AI Tech 2 min read

ChatGPT Work vs. Claude Cowork: Security Makes the Difference

Hands-on testing of two desktop automation tools reveals that while ChatGPT Work and Claude Cowork offer similar performance, they differ significantly in data security.

Tier 2 · sources 54% confidence Reviewed
Sources zdnet.com

Browser and desktop automation tools are becoming the new battleground for AI assistants, notably OpenAI's ChatGPT Work and Anthropic's Claude Cowork. Both solutions aim to help users automate tasks directly on their operating systems and local files. However, a recent real-world test highlighted core differences in user experience and, more importantly, data security between these two heavyweights.

Detailed Developments

During real-world testing with system files, both tools demonstrated impressive integration with desktop interfaces. Testers granted both ChatGPT Work and Claude Cowork access to document folders to perform office tasks and automated data analysis. Initial results showed that both AI assistants successfully completed basic requests, significantly reducing manual effort. However, the permission setup process and operational workflows of the two tools offered vastly different psychological experiences for users.

Technical & Technological Analysis

Technically, both ChatGPT Work and Claude Cowork utilize APIs that integrate deeply with the operating system to read and write data directly. They can recognize folder structures, open third-party applications, and execute simulated mouse and keyboard sequences. The main differentiator lies in permission control and security sandboxing. According to hands-on testing by ZDNet, ChatGPT Work interacts with system files with a higher degree of freedom, whereas Claude Cowork enforces stricter boundaries, mitigating the risk of the AI unilaterally modifying or deleting sensitive data outside the specified scope.

Expert Opinions & Insights

An testing expert from ZDNet noted that although their operational performance is comparable, ChatGPT Work's current mechanism feels less secure for users. Allowing a Large Language Model (LLM) unrestricted access to a file system without robust intermediate verification steps poses a major security concern. In contrast, Anthropic's Claude Cowork provides greater peace of mind, thanks to a design philosophy focused on 'constitutional AI'—a signature approach of the company.

Impact & Future

The trend of integrating AI agents deeply into desktop operating systems is bound to surge in the near future. For tech users in Vietnam, carefully weighing convenience against data security is crucial before entrusting personal data to these automation tools. The cautious approach of developers like Anthropic may well become an industry standard for future generations of AI assistants to ensure maximum user privacy.