Defenders have long leaned on a familiar rhythm: find vulnerabilities, release patches, deploy them, and move on. But new research points to a worrying reality—an AI-driven vulnerability surge is compressing timelines and stretching security teams beyond what monthly patch cycles can realistically handle.
Recent analysis by Rapid7 describes how the current environment feels less like normal operations and more like a stress test. Vulnerabilities are being disclosed at a higher volume, proof-of-concept code is appearing faster, and attackers are increasingly turning public information into operational access. In that setting, patching alone is no longer the reliable “get out of trouble” strategy it once was.
Why the old patch model is getting overwhelmed
According to the Rapid7 findings, the gap between what defenders need to manage and what they can actually triage is widening. The volume of high and critical vulnerabilities (CVSS 7 to 10) doubled from 4,268 in Q2 2025 to 8,539 in Q2 2026. In the same period, the number of newly exploited vulnerabilities rose by 8% to 40.
These figures matter because the ecosystem is not just producing more vulnerabilities—it is also moving faster. The result is a pressure effect: defenders face too many items to patch in time, while attackers shorten the window between discovery and exploitation.
Discovery and exploitation are not the same problem
Rapid7’s cyber intelligence leadership stresses an important nuance. Finding a vulnerability and exploiting it are separate issues. AI can assist attackers in both steps, but exploitation still depends on whether the target is reachable.
If a system is effectively shielded by multiple defensive controls—such as firewalls and other restrictions—an exploit may not translate into access. That’s part of why Rapid7 frames the situation as an exposure challenge, not merely a patch volume challenge.
AI also accelerates vulnerability creation
It’s tempting to think that patching breaks the cycle, but the source problem may keep regenerating. Rapid7 notes that vulnerability volume is unlikely to decrease because new applications keep shipping—often with new security flaws.
Even more concerning, the analysis points to the growing use of “vibe coding,” where AI helps generate code quickly. Rapid7 cites research involving vibe-coded financial applications that shared the same underlying vulnerabilities. The implication is that AI-assisted development can reuse old templates—and therefore repeat old mistakes.
In practical terms, this means new code may contain vulnerabilities that then show up in new AI-assisted scanning runs. So defenders do not just inherit a one-time backlog; they keep receiving new waves.
The exposure gap widens across modern attack surfaces
Another driver behind the patching problem is asymmetry. Attackers generally need only one weak spot, while defenders must protect a far larger set of assets. Traditional endpoints like laptops, desktops, servers, and firewalls are already part of that burden.
But the environment is changing fast. Organizations now rely heavily on APIs and the supply chain, and they depend on multiple vendors. That matters because the defender’s visibility is often less complete than the attacker’s ability to identify reachable paths.
Rapid7 describes this as a “widening gap between what’s disclosed and what any team can realistically triage.” Even if teams patch diligently, the attack surface may include paths that are hard to enumerate, difficult to segment, or simply unknown until something goes wrong.
“Holy Grail” vulnerabilities raise the stakes
Beyond sheer volume, Rapid7 highlights a rise in what it calls “Holy Grail” vulnerabilities. These are flaws that do not require credentials or user interaction. Such issues are dangerous because they reduce the attacker’s prerequisites.
Rapid7 reports a 9-point year-over-year increase in this type of vulnerability. In Q2 2026, they accounted for 25 of 40 exploited vulnerabilities. The key point is not only that exploitation is happening—it’s happening without the typical friction that slows intrusions.
From an attacker’s perspective, that can enable execution close to a device or product without needing authentication. For defenders, it raises the likelihood that a vulnerability will turn into real-world compromise before patching can catch up.
Nation-state pressure still matters
Rapid7’s report also highlights persistent activity associated with nation-state operations. The analysis references the so-called “axis of evil” set—commonly grouped as China, Russia, Iran, and North Korea (CRINK)—and notes distinct targeting patterns.
Russia is described as active primarily in Ukraine and against Ukraine’s supporters. Iran is portrayed as targeting the US and US allies. China’s activity is linked to Taiwan. North Korea, meanwhile, is described as targeting anything it can monetize.
Different motivations, different capabilities
Rapid7’s cyber intelligence leadership explains that nation-state advanced persistent threats are not necessarily more technically advanced than financially motivated criminal groups. The difference is motivation and resources.
Nation-state actors often require persistence for long-term espionage, and they may also have the budget and resources to develop sophisticated capabilities. In contrast, cybercriminal groups may primarily need access—something they can buy—then steal and exit.
This distinction reinforces why the patching mindset alone can fail: long-term actors can plan around timelines, while financially motivated actors can move quickly when profitable opportunities appear.
Ransomware continues to drive financial impact
Ransomware remains a major monetization method. In the reported quarter, the US is identified as the primary target by a wide margin. Germany ranks second, but the scale difference is stark: 881 victims in the US versus 91 victims in Germany in Q2 2026.
The analysis lists several of the most active ransomware groups in order: Qilin, The Gentlemen, DragonForce, Akira, and LockBit.
It also provides a breakdown of targeted sectors. Business services account for 23.5%, healthcare 22.0%, manufacturing 21.0%, technology 16.9%, and construction 16.6%.
While ransomware is not new, the combination of AI-assisted discovery and faster exploitation means the window between “known issue” and “active incident” can keep shrinking.
So what should defenders do instead?
Rapid7’s core message is straightforward: if the environment has changed, the strategy must change too. The report argues that a solution cannot be found through reaction alone. Instead of trying to patch everything on a predictable schedule, defenders should get ahead by reducing exposure.
The contrast between vulnerability disclosure numbers and the smaller set of exploited vulnerabilities suggests something critical. Not every disclosed flaw becomes a compromise. That difference implies that protective measures that limit reachability can be effective.
Stop focusing on CVSS alone
Rapid7 also calls out a common trap: over-relying on CVSS severity scores as the primary lens for prioritization. In response, the guidance is to focus on exposure—where the vulnerable component exists in the network and what would happen if an exploit succeeded.
Rather than assuming a monthly patch cycle will always work, defenders should treat new vulnerabilities as exposure-driven priorities. The practical questions shift to: where is it reachable, and what is the impact if it can be exploited from the attacker’s perspective?
In short, AI-driven vulnerability surge dynamics make timeliness essential, but timeliness starts with understanding reachability and reducing the paths that matter.
Making exposure reduction actionable
While Rapid7 emphasizes the concept, the underlying approach is practical: identify which parts of your environment attackers can realistically reach, then reduce those reachable options. This can involve strengthening network segmentation, hardening perimeter and internal access controls, and improving defensive visibility around APIs and third-party dependencies.
It also means treating patching as only one part of a broader exposure strategy. Even when patches are available, compensating controls may be needed to prevent exploitation during deployment delays or operational constraints.
When done well, defenders can shrink the gap between “vulnerabilities exist” and “attackers can use them.”
Conclusion: patching needs an upgrade
The research paints a clear picture. An AI-driven vulnerability surge increases both the number of disclosed issues and the speed at which exploits can move from idea to action. Meanwhile, modern architectures—APIs, vendor ecosystems, and evolving attack surfaces—make visibility and triage harder than before.
That’s why the report argues the traditional model of patching as the primary escape route is breaking down. The path forward is to prioritize exposure, understand where vulnerabilities can be reached, and reduce the reachable attack surface so that fewer disclosed flaws turn into real incidents.
Source: https://www.securityweek.com/ai-driven-vulnerability-surge-breaks-the-traditional-patching-model/
