Tide Foundation Releases Raziel — AI Security That Treats the...

Tide Foundation Releases Raziel — AI Security That Treats the Attacker as Already Root

Tide Foundation has released Raziel, an MCP server for developer AI assistants that audits systems on the assumption that an attacker is already inside, relying on the concept of emergent authority, where permissions are cryptographically computed only at the moment of a request and immediately disappear.

Compiled by Sergey KostenchukPublished 2026-09-03Updated 2026-09-18

2026-09-03 Safety

🛡 Raziel: AI Security That Treats the Attacker as Already Root

Tide Foundation has released Raziel, an MCP server for developer AI assistants. It scans 'in reverse': it assumes the hacker is already inside and builds a map of the blast radius. Under the concept of emergent authority, permissions are cryptographically computed only when identity, policy, and context align — and then immediately disappear.

🌍 Instead of a central secret store (vaults, MFA, EDR), authority is computed in parts and cannot be reassembled — reducing the cost of a single error in code or configuration. Security vendors are beginning to design products for agentic access from the start.

👤 Piloting AI agents? Raziel generates TideCloak integration playbooks and assesses the blast radius of your system. Remember: 'even root won't get anything valuable' — a vendor claim without independent verification.

Source 1: https://thenewstack.io/tide-raziel-emergent-authority/ Source 2: https://tide.org/agent-plugins

Sources