Gartner says frontier AI makes patch speed a vanity metric. Tolerable exposure time replaces it.
Gartner’s new research on frontier AI vulnerability management says the patch-speed metric most security teams report to their boards no longer measures anything useful.
Gartner analysts Rob McMillan and Alex Michaels put it bluntly in a research note published August 6: “Measuring ‘how fast we patch’ is a vanity metric; the only metric that matters now is ‘tolerable exposure time.'”
The note covers one of Gartner’s nine top cybersecurity trends for 2026: frontier AI models transforming vulnerability management. Its central argument runs against most vendor messaging. The real shock from frontier AI is not faster attackers. It is the volume of vulnerabilities that vendors and open-source projects will now find in their own code, and the flood of urgent updates that follows.
Gartner calls it a supply side shock, one that “permanently breaks traditional volume-based patching processes.”
I built this analysis from Gartner’s research note (G00859901) and paired it with category data from Gartner’s 3Q26 information security forecast (G00862059), which sizes the markets security teams will buy from as they shift to exposure management.
The bars in the hero chart show 2026 and 2030 end-user spending for four exposure-focused categories. Six tiles underneath summarize the note’s core findings.
Why frontier AI breaks volume-based vulnerability management
Gartner is direct about where the pressure comes from. Frontier models can compress attacker cycle times by automating reconnaissance and payload iteration, the analysts write, but “attackers have historically always been faster than defenders.”
The change is on the supply side. As vendors and open-source projects use AI to uncover latent flaws, organizations “will face a greatly heightened volume of urgent security updates.” The note expects those updates to arrive in multiple waves, timed in large part to when new security-focused frontier models become available to IT suppliers.
The exploit window is shrinking at the same time. Gartner says the gap between vulnerability disclosure and active exploitation “has collapsed to minutes because AI models can autonomously reverse-engineer patches into exploits.”
Gartner’s four core insights:
- Traditional vulnerability management was already misaligned with attacker timelines before AI. Without change, the gap is now unbridgeable.
- Mandates to “patch everything faster” move risk from successful attacks to operational outages and increase self-inflicted downtime.
- The bottleneck is testing and change management, not patching capacity. Each wave of AI-discovered vulnerabilities adds complexity and strains resources.
- Regulatory pressure for rapid patching can conflict with operational reality. Storytelling about real business impact becomes a necessary skill.
“Patch everything faster” moves the risk, it doesn’t remove it
The note treats blanket speed mandates as a risk transfer. Unrealistic demands to patch everything immediately “merely transfer the risk from malicious actors to insufficient IT testing, increasing the likelihood of self-inflicted downtime.”
Gartner frames enterprise risk management as an explicit choice between breach risk and outage risk. The CISO, infrastructure and operations leaders and other stakeholders need business leaders to formally accept the relative risks of patching-related downtime against attack-related downtime.
Faster patching still helps at the margins. The note says the gains are capped by maintenance windows, the need for bespoke fixes in complex software stacks and other constraints.
From volume SLAs to exposure SLAs
Gartner’s replacement is a shift from volume-based patching service-level agreements to exposure-based ones.
Static vulnerability characteristics, including CVSS scores and scanning results, are “still necessary, but it is no longer sufficient for prioritization of remedies,” the note says. Prioritization must rest on real exploitability and business context.
Gartner wants the most urgent vulnerabilities, “not necessarily the highest impact,” first in the queue, and mitigated within the organization’s maximum tolerable exposure time.
Accountability moves too. Product and platform teams are held to their exposure limits, with the flexibility to use virtual patching or network isolation when change management blocks an immediate code fix.
CTEM becomes the operating model
Gartner tells organizations to abandon volume-based patching and adopt continuous threat exposure management, or CTEM. Its five stages run from scoping and discovery through prioritization, validation and mobilization.
The note warns that “a failure to shift to a CTEM approach will result in continued patch deferral and therefore a slow but continuing escalation in risk profile.” Update complexity, and the time and collaboration needed to remediate, is likely to rise with each wave.
Gartner’s broader information security forecast points the same direction. Its 3Q26 report expects organizations to shift from reactive defense toward CTEM, with AI assistants automating event triage, false-positive reduction and code remediation.
The spending signal in Gartner’s security forecast
The categories that support exposure-based prioritization grow faster than the security market overall. Gartner’s 3Q26 forecast puts total information security spending at $247.5 billion in 2026 and $372.8 billion by 2030, a 10.8% CAGR in constant currency from 2025.
| Category | 2026 | 2030 | 2025-2030 CAGR |
|---|---|---|---|
| Threat intelligence | $3.1B | $6.1B | 19.0% |
| Network detection and response | $2.6B | $4.1B | 12.4% |
| Vulnerability assessment | $4.1B | $6.4B | 12.0% |
| Managed security operations | $15.6B | $22.7B | +$7.0B added |
| Total information security | $247.5B | $372.8B | 10.8% |
End-user spending in current U.S. dollars; CAGRs in constant currency. Source: Gartner G00862059 (September 2026).
Threat intelligence, the data CTEM uses to judge real-world exploitability, nearly doubles by 2030. Managed security operations adds $7.0 billion, the largest dollar gain among security services. I covered the full forecast in Gartner’s $247.5B security forecast makes post-quantum firewall refreshes the only upgrade through 2030.
Technical debt can’t wait for “later”
The note uses the AI-driven update surge to force a decision most organizations defer. Technical debt “generally falls into the ‘we will take care of it later’ bucket,” the analysts write. “‘Later’ must be ‘now.'”
CISOs guide the prioritization. CIOs, enterprise architecture and I&O teams have to execute the remediation in the short to medium term. Technical debt also appears separately in Gartner’s 2026 trends list, where it undermines organizational AI ambitions.
Use agentic application security testing selectively
Gartner recommends agentic application security testing, or AAST, for mission-critical applications. AAST can uncover business-logic vulnerabilities and multistep exploits by using context windows to reason through application logic.
The note pairs its reachability analysis, vulnerability prioritization and automated code remediation as the way to keep pace with AI-driven discovery volume.
There is a budget catch. Token consumption “is likely to drive an increase in application security budgets that many organizations will struggle to absorb under current assumptions.” Gartner’s advice is to use AAST selectively now and plan for the budget impact later.
Regulators want counts, CISOs need a story
Gartner says regulators, including the ECB and ESRB, already demand mitigation plans for systemic cyber risks caused by AI. One Gartner client reported pressure from a regulator to address a set number of vulnerabilities within 30 days.
The analysts read that as regulators not always understanding that a more surgical approach is sometimes needed “in order that one problem isn’t substituted for a worse problem.” In the short term, organizations may have to comply with both approaches in parallel.
Gartner expects regulation of AI in critical business processes to tighten. Third-party contracts will likely need vendors to disclose AI model usage and set contingencies for provider outages. Some frontier models may also face export controls or tighter access restrictions. The note cites Anthropic’s Fable 5 and Mythos 5 in 2026 as examples and advises building model-independent roadmaps.
Board talking points Gartner gives CISOs
Gartner’s Figure 1 gives CISOs three talking points for boards worried about Mythos, Daybreak and frontier AI risk:
- Foundational controls. Threat actors must still chain attacks. Core controls still provide adequate protection against advanced threats.
- Prioritization. A vulnerability does not equal exploitability. Environmental constraints, required privileges and mitigating controls make many flaws impractical to exploit.
- Accelerator. Defensive AI scales the defender’s advantage through automated code analysis, threat intelligence and incident response.
The strategic shift, in Gartner’s words, is from asking “how fast do we patch?” to asking “what is our tolerance for remaining exposed to a known vulnerability?”
Where this sits in Gartner’s 2026 trends
Gartner groups its top cybersecurity trends for 2026 into three themes: secure new frontiers, transform governance and normalize AI adoption. Frontier AI and vulnerability management sits under normalizing AI adoption, alongside AI-driven SOC solutions and the breakdown of traditional security awareness tactics.
What security leaders should do now
- Retire volume-based patch SLAs. Replace them with exposure-based SLAs tied to maximum tolerable exposure time, and adopt CTEM.
- Get formal risk acceptance. Have business leaders sign off on the tradeoff between patch-related downtime and attack-related downtime.
- Fund testing and change management. That is the bottleneck Gartner identifies, not patch deployment.
- Pay down technical debt now. Each AI-driven update wave makes legacy remediation harder.
- Pilot AAST on mission-critical apps. Budget for token costs before scaling it.
- Build model-independent AI roadmaps. Plan for export controls, access limits and provider outages.
Frequently asked questions
What does Gartner say about frontier AI and vulnerability management? Gartner says frontier AI is driving an unprecedented volume of vulnerability discoveries that makes volume-driven patching obsolete. It recommends exposure-based SLAs and continuous threat exposure management.
What is tolerable exposure time? Gartner defines the meaningful metric as how long critical vulnerabilities remain unmitigated, measured against the organization’s maximum tolerable exposure time.
Why not just patch faster? Gartner says blanket faster-patching mandates shift risk from attacks to self-inflicted outages, because testing and change management are the real bottleneck.
What is AAST? Agentic application security testing uses AI agents to find business-logic flaws and multistep exploits. Gartner recommends using it selectively because token consumption raises application security costs.
How I built this analysis
Quotes and recommendations come from Gartner’s research note G00859901, published August 6, 2026, as part of its Top Trends in Cybersecurity, 2H26 research. Category spending and CAGRs come from Gartner’s 3Q26 information security forecast (G00862059). The pairing of the two documents, the comparison table and the CTEM stage notes are my analysis.
Related analysis on this site:
- Gartner’s $247.5B security forecast makes post-quantum firewall refreshes the only upgrade through 2030
- Gartner’s AI security forecast exposes 162x services growth that still trails software 2 to 1 in new spending
- Top 6 cybersecurity trends from Gartner’s 2026 Security Forecast
- Gartner’s $812.6B app software forecast sends 38% of new spending to CRM as agentic AI targets $116B of legacy apps
This post is my personal analysis of Gartner’s research and does not represent my employer.
Sources
- Gartner, Cybersecurity Trend: Frontier AI Models Are Transforming Vulnerability Management Strategies, Rob McMillan, Alex Michaels, 6 August 2026, G00859901.
- Gartner, Forecast: Information Security, Worldwide, 2024-2030, 3Q26, 24 September 2026, G00862059.
