What's Included
Credential exposure scanning: a scan of the code repositories behind your AI tools and agents, up to a defined repository ceiling, focused on AI service credentials. Everything the scan surfaces is reported, classified, and prioritized.
Permission and access-scope review: what each sanctioned AI tool and agent can reach, what it retains, and who can invoke it.
Shadow AI discovery: identification of AI tool usage outside the sanctioned toolset, built from your existing network and identity logs, so adoption you can't see today becomes visible.
A findings report presented before any recommendation: evidence first, standing on its own
A clear next-step recommendation only after the findings themselves are established
Findings outside the audit's boundary that surface incidentally are flagged in the report, not investigated. An exposed credential doesn't get ignored because it wasn't the target.
What's Not Included
Remediation of findings: scoped separately
Application-by-application permission audits beyond AI tools and agents
Network infrastructure assessment
Formal penetration testing: routes to a specialized partner
Full codebase or single-system security review: see Code & Platform Audit
How this differs from our other audits: the Copilot Readiness Audit is a pre-deployment check specific to Microsoft 365. The Code & Platform Audit is a deep-dive on one existing system's architecture and code. The AI Security Audit covers the AI-specific exposure surface across the organization: credentials, permissions, and unsanctioned usage.
Timeline: 2 weeks, confirmed in scoping.
Price: fixed for the defined scope, set during scoping.
Pattern We See
Security firm GitGuardian's most recent annual audit of public code repositories counted nearly 29 million exposed credentials in 2025 alone, and found that 64% of the credentials leaked back in 2022 are still valid today. The same research found credential leaks tied specifically to AI services jumped 81% year over year. A credential doesn't need to be stolen to become a liability; it just needs to still work, years after everyone assumed it had been rotated.
This audit identifies exposure that already exists. For reducing the everyday, unintentional behavior that creates new exposure going forward, see Katalyst Enable™.
FAQ
No. It's bounded to your AI exposure specifically: the credentials behind your AI tools and agents, what those tools can reach and retain, and any AI use happening outside your sanctioned toolset. Broader network assessments and penetration testing are different disciplines, and we say so rather than stretch the scope.
We flag it. Findings outside the audit's boundary get noted in the report rather than investigated, because an exposed credential doesn't get ignored just because it wasn't the target.
The shadow AI discovery component works from your existing network and identity logs, so it shows what's actually in use rather than what policy says should be. Most organizations find the gap between the two wider than expected.
Two weeks, ending in a findings report presented before any recommendation. Evidence first; the next-step conversation comes after.
