RC RANDOM CHAOS

Agents Inherit Platform Gaps That Governance Can't See

· via The Hacker News

Original source

From Email-to-Case to the Boardroom: How Inherited Agents Break Governance Built for Software

The Hacker News →

Salesforce Agentforce agents receive data through standard platform features like Email-to-Case, which converts customer emails into case records. When chained together, these documented features create an unintended attack surface: external email routes through case records into agent prompts and back out. The vulnerability lies not in the agent model but in how the surrounding application handles untrusted data—a configuration problem that existed long before the agent arrived.

Enterprise agent governance has become fundamentally divorced from operational reality. Most organizations document policies and principles but cannot answer basic operational questions: how many agents are running, what systems can they access, or who controls each one. Standard security tools cannot see inside third-party platforms where agents actually live. Meanwhile, agent creation has spread across departments using inherited permissions, and agents themselves now spawn other agents without review. The City-Forum campaign, active since March 2025, exploited these exact gaps across Salesforce and ServiceNow instances.

Regulatory frameworks have moved faster than governance practices. The EU AI Act imposes penalties matching GDPR severity—up to €35 million or 7% of global turnover. Enforcement history shows that organizations get fined not for imperfect policies but for policies that exist only on paper and cannot be demonstrated operationally. Effective agent governance requires a bottom-up approach: inventory, ownership assignment, configuration visibility, and review history. Only this operational evidence satisfies both security practice and regulatory scrutiny.

Read the full article

Continue reading at The Hacker News →

This is an AI-generated summary. Read the original for the full story.