Cyber Risks · The Hacker News

How Ceros Gives Security Teams Visibility and Control in Claude Code

Summary: Beyond Identity has released Ceros, an AI Trust Layer that installs on a developer's machine alongside Anthropic's Claude Code and records every action the agent takes, including tool calls, MCP server connections, and the device's security posture at the moment of invocation. Ceros captures device context in under 250 milliseconds, signs each activity log entry with a hardware-bound cryptographic key, and evaluates policies at runtime before an action executes, blocking unapproved MCP servers and restricting tools such as Bash or file reads in paths like ~/.ssh/ or /etc/.

Bottom line: Claude Code runs with the full inherited permissions of the developer who launched it, so it can reach credentials, production systems, and sensitive data before any network-layer tool sees a request leave the device. The MCP servers developers connect casually are data access paths nobody reviewed, and the gap between what security teams assume is connected and what actually is tends to be significant.

Read the original at The Hacker News ↗