Most High-Severity Cloud Security Findings Are Downgraded Once an AI Agent Checks the Environment
Help Net Security, Tuesday, October 13th, 2026
An AI agent re-grading 9,967 HIGH cloud findings changed 75% of them, mostly downward, once it checked environment context.
Sola Security researchers built an AI agent that checks cloud security findings against the environment they live in, since vendor rules assign one static severity regardless of context.
Across 9,967 HIGH findings from two commercial platforms in eight production environments, it changed 75.1% of grades, moving 58.5% down and 16.6% up.
The largest downgrade pattern (32.3%) was inert or orphaned resources, while confirmed public exposure raised severity.
Raising a grade required more evidence than lowering it, about half the verdicts relied on evidence from outside the flagged resource, and read-only AWS probes validated the facts behind the verdicts.