Victor Hernandez Guzman

@victorh.dev

🔐 Authentication & Authorization | 🛡️ Security | 🌐 Distributed Systems | 🤖 Artificial Intelligence | 💻 Software Development

Stop securing AI agents with standard service principals, it breaks enterprise IAM. I propose a 3-part test to answer this: when does a workload become an autonomous agent? Don't grant static permissions to non-deterministic agents. idpro.org/from-non-hum... #IAM #CloudSecurity #AgenticAI

From Non-Human Identity to Agent Identity

Non-human identities routinely outnumber humans. However, AI introduces a new class that breaks the security assumptions of NHI.

idpro.org

The OpenID Foundation’s AIIM CG released A Protocol-Agnostic Threat Classification for Cross-Boundary AI Agent Authorization. It maps multi-gateway agentic federation across six key security dimensions, a must-read for IAM and security architects.

openid.net

New chapter: I've joined the Agent Control Standard (ACS) contributor group & supported the OWASP GenAI Security Project! AI agents act for us, but lack standard tracking/stops. ACS is a vendor-neutral control layer fixing this with runtime policies & OpenTelemetry. Want to help? DM me!

Agent Control Standard (ACS)

Widescale adoption of AI agents depends on trust, and trust requires transparency and control. Enterprises cannot rely on black-box agents operating across cloud, SaaS, on-premises and endpoint enviro...

genai.owasp.org