CoChat Launches AI Platform With Human Approval for Risky Agent Actions
Summary
CoChat says its platform gives organizations visibility into AI use and pauses agent actions it considers dangerous, such as exposing sensitive data or deleting files, until a user approves or rejects them.
Key points
- CoChat launched a platform intended to provide visibility and governance over enterprise AI use.
- The platform provides access to multiple AI models and supports shared chats, custom assistants, and autonomous agents.
- CoChat says it can pause an agent's action when its control layer identifies a potentially dangerous instruction, then request user approval.
- The article describes potential risks from unapproved AI tools, including autonomous actions and inaccurate or differing model responses.
- CoChat promotes comparing outputs from multiple models and collaboration among users as ways to improve visibility and identify possible misdirection.
Article Details
- Event Type
- Launch of CoChat, an AI collaboration platform designed to provide enterprise visibility and governance for AI tools and autonomous agents.
- Impact
- CoChat is described as providing a control layer that can pause potentially dangerous agent actions for user approval and enable teams to compare AI outputs. The article presents this as a way to address risks from unmanaged shadow AI, including inaccurate outputs, sensitive-data exposure, and deletion of data; it reports no confirmed incidents or measured impact.