Perplexity announces "Computer," an AI agent that assigns work to other AI agents

Perplexity has launched 'Computer,' an AI system designed to manage and execute tasks by coordinating multiple AI agents. Users can specify desired outcomes, such as planning a marketing campaign or developing an app, which the system breaks down into subtasks assigned to various models, including Anthropic’s Claude Opus 4.6 and ChatGPT 5.2. While this technology aims to streamline workflows and enhance productivity, it raises significant concerns regarding the autonomous operation of AI agents and the management of sensitive data. The emergence of such tools, alongside others like OpenClaw, highlights potential risks, including serious errors and security vulnerabilities due to unregulated plugins. For example, OpenClaw has been associated with incidents where it inadvertently deleted user emails, raising issues of user control and data integrity. Although Perplexity Computer operates within a controlled environment to mitigate risks, it still faces challenges related to the inherent mistakes of large language models (LLMs). These developments underscore the necessity for careful oversight and regulation in AI deployment to balance innovation with safety, as unchecked AI power can lead to harmful outcomes.

Why This Matters

This article matters because it highlights the risks associated with increasingly autonomous AI systems. As AI tools like Perplexity's 'Computer' become more capable, they may inadvertently lead to privacy violations, data misuse, or other unintended consequences. Understanding these risks is crucial for developing ethical frameworks and regulations that ensure AI technologies benefit society without compromising safety or privacy.