AI agent credentials live in the same box as untrusted code. Two new architectures show where the blast radius actually stops.

AI agent credentials live in the same box as untrusted code. Two new architectures show where the blast radius actually stops.

AI Agent Credentials Live in the Same Box as Untrusted Code: Two New Architectures Show Where the Blast Radius Actually Stops

Section 1 – What happened?

At the recent RSA Conference 2026, four keynotes from prominent tech companies - Microsoft, Cisco, CrowdStrike, and Splunk - converged on a critical issue in AI security. They emphasized the need to extend zero-trust principles to AI agents, which are increasingly being used in organizations. Microsoft's Vasu Jakkal and Cisco's Jeetu Patel highlighted the risks associated with AI agents behaving like "supremely intelligent teenagers" with no fear of consequences. CrowdStrike's George Kurtz identified AI governance as the biggest gap in enterprise technology, while Splunk's John Morgan called for an agentic trust and governance model. This consensus was echoed by Cisco's Matt Caulfield, who stressed the importance of continuously verifying and scrutinizing AI agent actions to prevent them from going rogue.

Section 2 – Background & Context

Seventy-nine percent of organizations already use AI agents, according to PwC's 2025 AI Agent Survey. However, only 14.4% of respondents reported full security approval for their entire agent fleet, as per the Gravitee State of AI Agent Security 2026 report. A CSA survey presented at RSAC found that only 26% of organizations have AI governance policies in place. This gap between deployment velocity and security readiness has been described as a "governance emergency" by CSA's Agentic Trust Framework. The increasing reliance on AI agents has created a pressing need for robust security measures to prevent potential risks.

Section 3 – Impact on Swiss SMEs & Finance

The implications of this issue are significant for Swiss SMEs and the finance sector. As AI agents become more prevalent, the risk of security breaches and data compromise increases. Swiss banks and financial institutions, which are already subject to stringent regulatory requirements, must prioritize AI security to maintain trust and protect their customers' data. The adoption of new architectures and governance models will be crucial in mitigating these risks and ensuring the secure deployment of AI agents.

Section 4 – What to Watch

The recent RSAC 2026 keynotes have sparked a critical conversation about AI security. As the industry moves forward, it will be essential to monitor the development and adoption of new architectures and governance models. The two companies that shipped architectures answering the question differently will be worth watching, as their designs may reveal where the real risk sits. Additionally, the Swiss government and regulatory bodies will need to consider the implications of AI security on the country's financial sector and SMEs.

Source

Original Article: AI agent credentials live in the same box as untrusted code. Two new architectures show where the blast radius actually stops.

Published: April 10, 2026

Author: louiswcolumbus@gmail.com (Louis Columbus)

---

Disclaimer: This article is for informational purposes only and does not constitute financial advice. Consult a licensed financial advisor before making investment decisions.