Tag · 3 articles
AI Agent Security
Oct 20, 2026 · 9 min readAI Agent Impersonation: How Prompt Injection Turns an Agent's Identity Into the Attacker's
Prompt injection does not steal an AI agent's credentials, it borrows them. Here is how the confused-deputy attack works and how to stop it with delegated tokens, audience binding, DPoP and gateway policy.
Mar 21, 2026 · 7 min readIETF AIMS Framework: How AI Agents Will Authenticate with SPIFFE, WIMSE, and OAuth 2.0
IETF AIMS (draft-klrc-aiagent-auth-00) defines how AI agents authenticate using SPIFFE SVIDs, WIMSE workload identity, and OAuth 2.0 — replacing dangerous API keys with short-lived, attested credentials. 8-layer framework explained for developers.
OAuth 2.1
Feb 21, 2026 · 6 min readMCP OAuth 2.1 Authentication: How AI Agents Securely Connect to Tools
MCP OAuth 2.1 authentication explained with the Discovery Trifecta, mandatory PKCE, token audience binding, and IdP compatibility. Complete guide for securing AI agent connections.