I let ChatGPT Work and Claude Cowork loose on my files - only one made me nervous

Editor's take

OpenAI's ChatGPT Work and Anthropic's Claude Cowork both demonstrated desktop automation capabilities, but Claude's approach to file access and processing evoked less user apprehension. This distinction highlights a critical tension in AI development: balancing powerful functionality with user trust and data security. As LLMs become integrated into daily workflows, the perceived safety and transparency of their operational methods will directly influence adoption rates and competitive positioning.

The key differentiator lies in how each model handles sensitive user data. Claude's more cautious, perhaps sandboxed, interaction with local files suggests a deliberate design choice prioritizing privacy, which is paramount for enterprise adoption. ChatGPT Work's approach, while potentially more direct, raises immediate concerns about unauthorized data exposure. Future developments will likely involve refined access controls and clearer user consent mechanisms for AI assistants interacting with personal or company data.