Incident sheet · /security

Filed
Read
5 min · 934 words
Severity
Routine

Continuous security assurance still needs skilled people

ISACA’s 2026 survey finds 58% of security teams understaffed, pushing many toward annual risk reviews even as AI adoption accelerates and human oversight lags.

Continuous cybersecurity assurance takes more than automation
FeedContinuous cybersecurity assurance takes more than automation — pixelcreatures / pixabay

Event log

  1. ISACA’s State of Cybersecurity 2026 survey covered more than 1,800 security professionals.
  2. 58% reported understaffed security teams; the median team had eight people.
  3. 39% of understaffed teams conduct cyber-risk assessments only once a year.
  4. 35% cite AI or automation as their main way to close technical skill gaps, up from 23% a year earlier.
  5. FedRAMP 20x requires at least six months of history for every applicable Key Security Indicator for Class C certification.

By Olivia Hart

Fifty-eight percent of security teams are understaffed, and the median team has just eight people. Among understaffed teams, 39% conduct cyber-risk assessments only once a year; overstaffed teams assess more often. Cadence tracks labor capacity, not discipline.

ISACA’s State of Cybersecurity 2026 survey of more than 1,800 security professionals makes the constraint clear: time to run an assessment ranks as the leading obstacle at 35%, followed by insufficient staffing at 29%.

Hiring will not close the gap quickly. Forty-nine percent of respondents report unfilled security roles, and 24% say a non-entry-level role takes more than six months to fill.

Automation provides the immediate response: 35% now name greater reliance on AI or automation as their main way to close technical skill gaps, up from 23% a year earlier. ISACA’s figures nevertheless show why automation cannot run unattended.

The controls require a rhythm

Harshad Sadashiv Kadam’s analysis begins with NIST’s machine-readable OSCAL release of the SP 800-53 Rev. 5 catalog and baselines. The moderate baseline selects 287 controls.

His manual filter — NIST publishes no such mapping — narrows those to roughly two dozen controls a network policy plane can enforce or evidence: boundary protection, flow enforcement, remote access, internal connections, information exchange, network disconnect and traffic monitoring.

  • SC-7 Boundary Protection
  • AC-4 Information Flow Enforcement
  • CA-9 Internal System Connections
  • SI-4(4) Inbound and Outbound Communications Traffic

Six of the filtered controls carry an organization-defined frequency or time period. SI-4(4) has two.

A filled-in frequency changes the obligation. The organization no longer asserts a state; it asserts a rhythm, and a rhythm needs a record spanning time.

ISACA’s updated IT Audit Framework now includes continuous assurance, reflecting a move from periodic review toward ongoing insight, according to Mary Carmichael. FedRAMP 20x goes further: a provider seeking Class C certification must supply at least six months of history for every applicable Key Security Indicator — a trend, not a snapshot.

Tools and reviewers

Continuous controls monitoring platforms, OSCAL-native compliance tools and cloud configuration history can retain results and test them against a committed interval. Cloud providers publish reference architectures for the same purpose.

The survey suggests tools are not the main barrier: their cost and lack of the right tools each draw 18%, behind time and staffing. Fifty-five percent report an underfunded security budget. Building a pipeline also consumes engineering time for connections, intervals and integration maintenance — time an eight-person team already spends on assessments.

Overstaffed teams have room to automate and build. Teams stuck in an annual cycle have the least capacity to escape it.

FedRAMP’s guidance to 20x assessors is blunt: “Automation is not proof by itself.” Assessors must test that validation runs at the stated cadence, covers the full scope and returns consistent results. They also have to scrutinize evidence from tools nobody can explain.

Someone still must understand the pipeline well enough to vouch for it. A missed run must appear as a gap rather than silence, scope must not drift, and the interval checked must match the security plan’s commitment. Automated evidence carries only as much trust as the people who can explain how it was produced.

AI automates routine security tasks at 40% of enterprises, up 12 percentage points in a year. Only 8% run AI-related incident response exercises regularly. LLM SecOps is the second-largest skill gap at 45%, up from 33%. The report’s conclusion is direct: “Adoption is outpacing readiness.” AI has not reduced the need for skilled practitioners or human oversight.

ISACA’s audit evidence discussion now asks whether a submitted artifact can be authenticated. A machine-generated series with clear provenance answers that question; a screenshot does not. The series holds up only if someone can explain the machine behind it.

Automation changes who can verify the evidence

Forty-two percent of respondents manage staff with fewer than three years of experience. The best reviewers learned to spot missed scans, shrinking scope and misleading exports by gathering evidence manually. Automate that work without teaching it, and the capable reviewer pool shrinks too.

What does this change for an eight-person team?

Start with controls carrying frequency parameters. Retain each result by control and timestamp, and store the committed interval beside it. That pairing makes a missed check visible, because a missed scan leaves no record of its own. It is a narrow, affordable place to begin.

Treat the pipeline as an assessed system, not merely an assessment generator. Document how it works, test it against its claimed cadence, and make sure more than one person can explain it. Use it to teach junior staff what the evidence means, not just where it lives.

FedRAMP’s 20x pilot lessons point the same way: trending validation data over time builds confidence in a provider’s security program. Not the scan. The trend.

Kadam is a senior infrastructure security engineer at Indeed, where he architects zero trust, SASE and edge-native security across multi-cloud infrastructure at enterprise scale. He builds open-source tools including MCP Threat Trap, KubeTrap and FlareGuard, speaks at security conferences on AI security, deception engineering and edge infrastructure, and holds the CISM certification. He serves on ISACA’s Emerging Trends Working Group and Academic & Workforce Development Advisory Group and is founding chapter lead of AI Security Engineers Austin.

ISACA’s survey locates the real constraint in evidence collection and the human capacity to vouch for it. Build the record. Then keep the people who can read it.

via techtarget.com (Source)

Filed under

  • continuous-security-assurance
  • cybersecurity-automation
  • security-staffing
  • ai-security
  • nist-oscal
Share this article:

More from Olivia Hart

Olivia Hart

Show full bio

News editor covering money, funding and market shifts around enterprise information technology and the leaders who run at CIO Monitor.

207 articles

The newsroom digest

One email a day with the stories that matter. No noise.

By subscribing you agree to the privacy policy.

Cross-references

‹ Previous articleNext article ›