How CISO Governance Is Changing in 2026
A CISO's job used to be reasonably bounded: protect the environment, manage the security team's own risk. That boundary is dissolving. Boards now expect CISOs to speak to AI governance, data governance, and enterprise resilience — while the technical debt sitting underneath most environments is quietly blocking the AI initiatives those same boards are pushing for. That's the territory covered by "Transform Governance," one of three themes in a 2H26 cybersecurity trends update from Gartner®, and it breaks down into three specific, practical problems: technical debt that blocks AI adoption, a CISO remit that's expanding faster than CISO authority, and AI agents employees are building faster than anyone can govern them.
This post breaks down what's changing in each of those three areas, and what actually closes the gap between "the board expects oversight" and "the security team can deliver it."
What Does "Transform Governance" Mean?
"Transform Governance" is one of three themes Gartner uses to organize its top cybersecurity trends for the second half of 2026, alongside "Secure New Frontiers" and "Normalize AI Adoption." Where "Secure New Frontiers" is about visibility into new categories of risk — AI supply chains, machine identity — "Transform Governance" is about how CISOs are expected to demonstrate control over risk they don't fully own: cross-functional domains like privacy and AI governance, and decisions employees make on their own without a security review.
The throughline across all three trends in this theme is evidence. A CISO asked to speak credibly to the board about AI governance, or to prove technical debt is actually getting remediated, or to show that ungoverned AI agents are under control, needs something better than a point-in-time audit or a survey employees mostly ignore. They need current, queryable proof.
Get the full report This post covers the "Transform Governance" theme from Gartner's "Top Trends in Cybersecurity — 2H26." Download the full report for all three themes, including how the AI attack surface is expanding and how frontier AI is reshaping vulnerability management.
Why Is Technical Debt Becoming an AI Governance Problem?
Technical debt — the accumulated shortcuts, unpatched systems, and undocumented dependencies most environments carry — used to be a maintenance conversation. It's now a governance conversation, because the same debt that engineering teams have deferred for years is exactly what stalls AI initiatives once the business tries to move fast on top of it.
49% of cybersecurity leaders cite technical debt as a major barrier to executing their top functional priorities, according to Gartner's 2026 CISO Role Agenda Poll. The mechanics are straightforward: business pressure to deploy AI pushes teams toward unstructured data repositories and legacy systems that were never built with AI-ready controls, producing problems like data oversharing and unauthorized access. At the same time, AI-accelerated vulnerability discovery is surfacing far more issues in that same legacy footprint than teams can patch through manual processes.
Why can't CISOs just own technical debt remediation directly?
Because most of it isn't security's debt to remediate. Data debt belongs to data owners, software debt belongs to engineering, infrastructure debt belongs to IT — security's job is to make the risk visible and quantifiable, not to become the team that fixes every legacy system enterprise-wide. Taking on direct ownership of debt remediation is how a security function ends up overcommitted and under-resourced.
What to do about it:
- Inventory in-flight and short-term initiatives, and rank them by how much they reduce risk from AI data access and AI-accelerated vulnerability discovery — not by how "modernization-adjacent" they sound.
- Build cross-functional partnerships with engineering, data, and enterprise architecture teams to identify the legacy infrastructure and unmanaged data repositories actually stalling AI initiatives.
- Reframe technical debt remediation in front of business leaders as an AI enablement issue, not routine maintenance — that reframe is usually what gets it funded.
- Track debt as a connected, queryable inventory — which systems are legacy, which lack required controls, which hold sensitive data — instead of a static spreadsheet that's out of date the week it's built.
Why Are CISOs Being Asked to Orchestrate Risk They Don't Own?
The CISO role is expanding to cover data governance, AI governance, privacy, and business continuity — driven by the shift from cyberprotection to cyber resilience, and by urgent pressure to secure AI adoption. That expanded remit raises the CISO's visibility with the board. It also raises the odds of being blamed for a failure in a domain the security team was never resourced to control.
93% of nonexecutive directors view cyber risk as a threat to shareholder value, and 98% believe cyber risk will increase over the next two years, per Gartner's 2026 Board of Directors Survey. Boards are paying closer attention to cyber risk than ever — which raises the stakes for how a CISO demonstrates oversight of the domains now nominally under their remit.
What does it mean to lead through influence instead of ownership?
It means treating adjacent domains — privacy, AI governance, data governance — as things the CISO coordinates and reports on, not things the security team directly operates. Practically, that requires the ability to show, with current evidence, what's covered and what isn't across those domains, without security having to be the team that physically implements every control.
That's an evidence problem before it's an authority problem. A CISO who can show the board a current, queryable answer to "what AI governance controls exist and are they actually enforced" is in a fundamentally different position than one relying on a quarterly attestation cycle. Continuous, cross-functional control testing — evaluated against live systems and configurations rather than documentation — is what makes that kind of answer possible without the security team taking on direct ownership of every domain it touches.
What to do about it:
- Achieve objectives through influence and coordination with peers — CIOs, chief risk officers, chief data officers — rather than direct task ownership. Resist absorbing new responsibilities without the resources to deliver on them.
- Center cyber resilience, not just prevention, as the outcome that matters to the board: minimizing business harm from a successful attack alongside ongoing prevention work.
- Build board-level cybersecurity competence proactively, using the same shareholder-risk framing boards are already applying to cyber.
- Replace point-in-time audits of cross-domain controls with continuous testing, so "orchestration" comes with evidence attached instead of resting on trust.
Why Does Agentic AI Need Program-Level Oversight?
Employees are building AI agents faster than security teams can track them — often with no-code and low-code tools, embedded inside existing enterprise software or run as informal, stand-alone automations. Beyond expanding the attack surface, this creates a specific governance problem: agents operating in the background with more access than anyone intended, unmonitored, and frequently abandoned after the person who built them moves on.
Only 16% of organizations have successfully enforced workflows to govern employees' use of third-party AI applications, according to Gartner's 2026 Surviving the Agentic AI Era Survey — even as 67% of organizations have observed unsanctioned, employee-driven use of GenAI agent platforms. That usage is frequently hidden: a separate industry survey found roughly a third of employees using generative AI at work keep that usage concealed, which makes discovery a real problem, not just a policy one.
Why can't existing IT policy cover this?
Because most agentic automation doesn't go through a procurement or IT review process at all — it's built directly by employees connecting existing tools to internal APIs, in pursuit of automation value the business is actively pushing for. CISOs can't slow that down without fighting the same AI adoption pressure driving the technical debt problem above. What they can do is treat agent governance as a program to run continuously, not a policy to publish once.
What to do about it:
- Build or acquire the capability to control AI usage, secure AI-to-tool communication, and manage identity for agents, rather than relying on a single point solution for any one piece.
- Prioritize data security and access management for agents earliest in their life cycle, since most prototypes never reach production — securing every early pilot with full rigor is often not where the risk actually concentrates.
- Extend discovery beyond simple visibility: distinguish sanctioned from unsanctioned agents, and surface intent, not just existence.
- Build a framework for categorizing agents by the sensitivity of the data they touch and the level of autonomy they hold, so oversight effort goes where the risk actually is.
- Treat "does this agent have a documented owner and an enforced access policy" as a control to test continuously — the same way you'd test any other cybersecurity control — rather than a one-time registration checkbox.
Why Continuous Evidence Is the Common Thread Behind All Three
Technical debt, cross-domain orchestration, and agentic AI oversight look like three separate governance problems, but they all fail the same way: with a static snapshot standing in for a live answer. A spreadsheet of technical debt goes stale the day after it's built. A quarterly attestation for a cross-domain control says nothing about what's true today. A one-time agent registration doesn't catch the agent an employee spun up last week.
JupiterOne's Continuous Controls Monitoring (CCM) capability is built to answer these questions continuously rather than periodically, evaluating controls against live asset and configuration data instead of point-in-time documentation. It sits on top of JupiterOne's graph-native Cyber Asset Attack Surface Management foundation, where assets, identities — including AI agents — and their relationships are modeled together, so a control gap doesn't just say "this control is failing," it shows exactly which systems, identities, or data are affected.
In practice, that turns governance questions that used to require a manual audit into something closer to a standing query: which systems still carry the technical debt flagged last quarter, which AI governance controls are actually enforced right now, which AI agents lack a documented owner. That's the difference between reporting governance status to the board and demonstrating it.
This is the second post in a series on Gartner's 2H26 cybersecurity trends. The first post covers the "Secure New Frontiers" theme — how AI agents and third-party models are reshaping software supply chain security and machine identity: read it here.
FAQ: AI Governance Basics for 2026
What does "Transform Governance" mean in Gartner's 2H26 report? "Transform Governance" is one of three themes in Gartner's 2H26 cybersecurity trends update, covering how CISO governance responsibilities are expanding to include technical debt remediation, cross-domain risk orchestration, and oversight of employee-built AI agents.
What is technical debt in a cybersecurity context and why does it block AI adoption? Technical debt refers to accumulated data, software, and infrastructure issues — unpatched systems, unstructured data repositories, undocumented dependencies — that organizations have deferred fixing. It blocks AI adoption because AI initiatives typically require clean, well-governed data and modern infrastructure, and legacy debt introduces risks like data oversharing when AI tools are deployed on top of it.
What does it mean for a CISO to lead through influence instead of direct ownership? It means the CISO coordinates and reports on cross-domain risk — privacy, AI governance, data governance — rather than having the security team directly own and implement every control in those domains. It relies on the CISO being able to show current, credible evidence of what's covered across those domains without taking on direct accountability for them.
Why does agentic AI need program-level oversight instead of just an IT policy? Because most AI agents are built directly by employees using no-code or low-code tools, bypassing procurement and IT review entirely. A published policy doesn't govern automation nobody has to check in with IT to create — it requires ongoing discovery, categorization, and continuous verification that each agent has a documented owner and enforced access controls.
How does continuous control testing differ from point-in-time compliance? Point-in-time compliance evaluates controls at a single moment — an audit, an attestation, a survey — and the result goes stale as soon as the environment changes. Continuous control testing evaluates controls against live asset and configuration data on an ongoing basis, so the answer to "is this control working" reflects what's true right now, not what was true at the last review.
