---
title: "What the SANS 2026 AI Survey Actually Tells Us"
description: "Every security survey released in 2026 has told roughly the same story: AI adoption's moving fast while governance crawls behind, and executives see things differently from practitioners. The..."
url: https://kaynemcgladrey.com/blog/what-the-sans-2026-ai-survey-actually-tells-us/
date: 2026-07-20
modified: 2026-07-20
author: "Kayne"
image: https://kaynemcgladrey.com/wp-content/uploads/2026/07/compare-and-contrast.webp
categories: ["Blog"]
type: post
lang: en
---

# What the SANS 2026 AI Survey Actually Tells Us

![SANS 2026 SANS AI Survey Insights 070926 cover](https://kaynemcgladrey.com/wp-content/uploads/2026/07/SANS-2026-SANS-AI-Survey-Insights-070926_cover-1024x769.webp)

Every security survey released in 2026 has told roughly the same story: AI adoption’s moving fast while governance crawls behind, and executives see things differently from practitioners. The question worth asking isn’t whether this pattern holds. It’s whether any single report actually adds insight beyond repeating what we already know.

Matt Bromiley’s July 2026 [SANS AI Survey](https://www.sans.org/white-papers/2026-sans-ai-survey-insights) Insights landed on my desk on the weekend with a sample of 536 security practitioners and 57 CISOs. Its headline finding was a nice change of pace: **78%** of organizations have experienced confirmed or suspected **AI-enabled attacks** (according to practitioners), but only **16%** of leaders shifted priority toward **defending against them**. That gap between threat awareness and resource allocation stands out amid the familiar chatter about adoption rates and governance shortfalls.

- Adoption jumped from 50% to 78% year-over-year in security strategy, but just 27% described their production deployment as mature
- 63% of practitioners reported significant AI shortcomings in threat detection and response, up from 45% the prior year
- Leadership claimed formal AI risk programs more often than practitioners could confirm: 50% versus 36%

The data paints a picture of rapid growth meeting operational friction, with tools scaling faster than they deliver measurable results.

I often cover reports here on the blog, and multiple reports released in 2026 have pointed out the same structural problems:

- **Bitdefender’s **[**Cybersecurity Assessment 2026**](https://www.bitdefender.com/content/dam/bitdefender/business/campaign/cybersecurity-assessment-2026/Bitdefender_Cybersecurity_Assessment_2026.pdf) found managers reporting 57.8% full AI visibility while practitioners sat at 45.9%. Same gap, different metric
- **Deloitte’s **[**State of AI in the Enterprise**](https://www.deloitte.com/us/state-of-ai) (January 2026) showed 74% of companies planning agentic AI deployment within two years, but only 21% had mature governance models
- **Thomson Reuters’s **[**Future of Professionals Report 2026**](https://institute.thomsonreuters.com/en/institute/future-of-professionals-2026/report) tracked 34% of professionals using shadow AI.
- **IBM’s **[**2026 Tech Leader Study**](https://www.ibm.com/thought-leadership/institute-business-value/en-us/c-suite-study/cxo) counted 54 average AI agent incidents per organization last year, 17% classified as high severity
- **Kroll’s **[**AI Innovation Surges, Security Gaps Persist**](https://www.kroll.com/en/newsroom/ai-innovation-security-lag-kroll-research) reported 76% of organizations experienced AI-related security incidents over two years, with 27% seeing costs exceed $1 million

Across multiple research groups with different sampling frames, we’re seeing the same core story. Adoption outruns oversight, leaders see more visibility than practitioners experience, and bad things happen when AI touches production systems without sufficient controls.

SANS breaks from this pack on three dimensions that matter.

**First**, it stays focused on security operations rather than general enterprise AI. Deloitte covers agentic AI broadly. IBM surveys C-suite technology leaders on scaling agents. Thomson Reuters examines professional work across industries. SANS zeroed in on 536 people who actually run security workflows, where the 78% adoption rate reflects AI embedded in detection and response, not HR automating recruitment emails.

**Second**, SANS provides temporal change data while most reports offer a snapshot. Adoption climbed 28 percentage points in one year. Shortcomings grew 18 points. That trajectory matters more than any single measurement, because when problems widen rather than stabilize, the narrative shifts from “this is challenging” to “this is getting worse.”

**Third**, SANS goes further by linking incident exposure directly to strategic response. Leaders aren’t standing still: 63% shifted priority toward using AI for defense over the past year, a move the report itself calls reasonable, given AI’s real leverage in detection and response. But only 16% shifted toward defending against AI-enabled threats, even as 78% of organizations already face them and 95% of leaders believe attackers are using AI. The defensive bet is sound. The gap is that adversarial AI is still being treated as a horizon item when it’s already operational.

[DigiCert’s AI Trust Pulse](https://www.digicert.com/content/dam/digicert/pdfs/report/ai-trust-pulse.pdf) offered a different angle, measuring governance hygiene like inventorying AI systems, dedicated security budgets (57%), processes to revoke compromised systems (86%), and liability evaluation (nearly 90%). That’s genuine coverage of whether controls exist. But existence isn’t efficacy. SANS asks whether the security tools actually work in practice, with 63% saying they fall short, which is a different and harder question than whether an organization has drawn up its inventory.

Bitdefender noted 53% of respondents believe AI helps attackers more than defenders, and only 17.5% view AI attacks as “mostly industry hype”. The skepticism is fading. But SANS’s contribution sits in what comes next.

We’ve established AI poses risks. We’ve measured the gaps. The real question is whether anyone acts on what they know. Here’s where vendor influence becomes relevant. Nearly every report cited here carries some commercial footprint. Deloitte sells advisory services, Kroll markets risk solutions, Bitdefender makes security products, IBM pushes cloud infrastructure, DigiCert distributes certificates. And SANS’s survey is similarly sponsor-funded, with 18 named sponsors on the cover and sponsor page.

- Governance lag appears across every study
- Shadow AI shows up repeatedly
- The leadership-practitioner split confirms itself through multiple lenses

SANS didn’t discover these tensions; it measured how quickly they’re intensifying within cybersecurity specifically. When 78% of organizations face attacks and only 16% of leaders prioritize defending against them, the problem isn’t awareness – 95% of leaders know attackers are already using AI. Nor is it inertia; leaders are reallocating, just toward AI-assisted defense rather than adversarial-AI protection. That mismatch between what they know and where the counterweight lands is the real vulnerability, not any technical limitation in AI detection itself.

SANS’s value sits in security-specific granularity paired with year-over-year trajectory. Every other report captures a moment. SANS captured momentum. In an environment where AI deployment accelerates month to month, understanding direction matters more than documenting position. A year ago, 50% of security teams used AI. Now 78% do. A year ago, 45% reported AI shortcomings. Now 63% do. Those aren’t rounding errors. They’re directional signals that demand strategic response rather than tactical adjustment.

Don’t treat the governance gap as abstract. Track it quantitatively. Measure whether incident exposure drives priority shifts. If it doesn’t, the gap will keep widening regardless of what vendors promise. Until leaders connect threat awareness to budget allocation, the poisoned wells will keep flowing. The “pure springs” remain theoretical. SANS didn’t solve either problem. It just made the distinction harder to ignore.