Enterprise AI adoption on macOS fleets has outrun the AI governance frameworks meant to contain it. Jamf, the Apple device management vendor, surveyed 687 IT and security leaders managing Apple-first environments. More than one in five have already lost money or experienced a cyberattack tied to AI tool use. Deeper AI integration correlates with higher incident rates, yet governance ranked only third on respondents’ priority lists.
- 72.9% of macOS-based organizations have already deployed AI, with another 20% actively exploring deployments.
- Organizations with deeply integrated AI programs reported incidents at a 27.1% rate, versus 19.4% among early-stage adopters – a 40% gap that widens as adoption matures.
- 81.7% of respondents have either experienced an AI-related incident or expect one, while governance placed third among stated priorities.
- Shadow AI, agentic AI, vendor sprawl, and unpredictable usage costs are the four recurring challenges driving exposure across Apple-first fleets.
Jamf’s 687-Leader Survey: AI Depth, Incident Rates, and the AI Governance Gap
The Jamf survey surfaces a pattern that security teams in deep AI deployments are already living: the further an organization integrates AI, the more incidents it accumulates. Among organizations still exploring AI, the incident rate was 19.4%. Among those that have deeply integrated AI into their operations, the rate climbed to 27.1%. That 40% differential reflects a standard exposure curve. What stands out is where governance sits on the priority stack. Automating IT management ranked first at 44.4% of respondents. Deploying AI productivity tools ranked second at 41.0%. AI governance placed third at 36.7%. AI security improvements ranked fifth. Organizations are accelerating into the deepest end of the incident curve while deprioritizing the controls that close the gap.
The biggest named threat in the survey is shadow AI – employees’ use of unapproved and ungoverned AI tools. When IT teams lack visibility into which AI systems are in use, Jamf noted, “that lack of visibility makes security and governance difficult, if not impossible.” Agentic AI compounds the problem. IT and security leaders told Jamf they struggled to deploy AI agents without putting data at risk. With appropriate code-write permissions, an AI agent can silently add insecure code to or remove necessary code from production repositories. Vendor sprawl adds a third pressure: the pace of AI integration into existing enterprise products makes it hard for IT teams to vet each tool before it reaches end users. CSI’s earlier coverage of the shadow AI governance gap found similar dynamics across enterprise environments.
Why Governance Lags Behind AI Adoption: Speed, Embedded Features, and Split Ownership
Jamf frames the problem as an efficiency challenge. That framing undersells the structural issue the data reveals. AI adoption on macOS fleets is driven by business-unit demand and by AI features shipping as updates inside already-approved productivity suites. When an AI model arrives as a feature update, IT does not vet it before it reaches users. The “shadow AI” category therefore includes AI features embedded in sanctioned software that IT never evaluated for data-handling behavior.
The usage-based cost model adds a dimension that typically sits outside security teams entirely. Jamf found that many organizations lack visibility into how many paid AI licenses they hold and which tools deliver value. Security teams watch data-access logs. Finance watches the billing console. Neither team has the full picture of AI exposure. That ownership gap explains why 81.7% of respondents have experienced or expect an incident despite the high-confidence adoption rates. The CSC CISO Survey also found that AI security optimism outpaces actual domain controls across enterprise programs.
Closing the Gap: Three Steps From Jamf’s Survey Data
Jamf’s recommendations address the incident-rate differential directly. Each step targets a specific exposure the survey data identified as driving the gap between early-stage and deeply-integrated programs.
Audit AI tool inventory before tightening policy – Jamf recommended expanding visibility through regular audits as the first step. Governance decisions made without an accurate tool inventory produce incomplete policies. Teams cannot restrict what they have not counted, and the shadow AI threat begins with an uncounted population of tools.
Enforce software access controls, not user-behavior rules – Jamf specifically recommended software governance over user governance, through enforced data-access policies. User-behavior rules are hard to scale and generate friction with productivity goals. Enforcing which tools can access enterprise data systems closes the exposure at the software layer. It does not require individual behavior change from every employee.
Start AI governance at the procurement stage – Organizations that deploy AI productivity tooling first and retrofit controls later face the steepest climb. They are already at the 27.1% cyberattack and incident rate while trying to govern tools already in production. For the 20% of macOS-based organizations still exploring AI, the window to build AI governance before deployment is still open.
Join our LinkedIn group Information Security Community!
