For three decades, vulnerability management operated thanks to a cushion: the time gap between when a flaw was discovered and when someone could develop a way to exploit it. The approach was simple enough; prioritize by severity, schedule the fix, confirm it, and proceed. That cushion made the system work.
Now, that cushion no longer exists.
AI hasn't slowed your team down. It altered the dynamics on the opposing side, shrinking the timeframe from discovery to exploitation from months to hours. And the unfortunate reality for defenders is that a methodology reliant on having time to spare cannot function without it.
AI Transformed Vulnerability Discovery Into a Numbers Game
In its May 2026 update, Anthropic stated that it and roughly 50 collaborators employed Claude Mythos Preview to uncover more than 10,000 high- or critical-severity vulnerabilities in essential software within a single month.
Earlier statistics were equally striking.
When focused on Firefox, the restricted Mythos model generated 181 functional exploits, compared to just 2 from the earlier frontier model. It uncovered vulnerabilities across all major operating systems and browsers, including an OpenBSD flaw that had remained hidden for 27 years.
As of this writing, over 99% of what it discovered remained without patches.
An AWS threat-intelligence report from February 2026 illustrates the other side: no zero-days required, only weak credentials, automated through a custom MCP server running offensive tools independently. AWS identified more than 600 devices across over 55 countries; independent researchers noted that the actor's logs queued up 2,516 devices spanning 106 countries.
In either scenario, the landscape has clearly shifted. What previously demanded specialized expertise now operates at machine speed and scale.
The Window for Weaponizing Vulnerabilities Has Also Disappeared
Defenders used to enjoy months between a CVE becoming public and its first confirmed exploitation in the wild, a period known as time-to-exploit (TTE).
That window has now closed abruptly.
Zero Day Clock places the 2026 average at roughly 24 hours, a drop from about 53 days in 2024.
Breach data corroborates this as well.
Verizon's 2026 DBIR attributes 32% of initial-access techniques to exploiting vulnerabilities and anticipates that figure will rise, as AI coding assistants now make exploit development, porting tools to new languages, and finding new flaws all accessible to attackers who previously lacked these capabilities.
Telling Teams to Patch Faster Is Like Asking a Freighter to Stop on a Dime
The industry's automatic response is to patch more quickly. Regulators are formalizing this: numerous regulations now suggest same-day fixes for certain critical vulnerabilities. Boards anticipate it. Executives demand it.
However, remediation is not a simple toggle. Patches must pass regression testing, wait for change windows, require approvals, and honor existing uptime and compliance obligations. Taking production offline to outpace an exploit simply results in a different kind of outage.
And the data indicates that everything is trending in the wrong direction.
The Verizon 2026 DBIR tracked over 13,000 organizations:
Median fix time for known-exploited vulnerabilities: 43 days, up from 32 the previous year
Percentage that were fully patched: dropped from 38% to 26%
When offense operates in hours and remediation takes weeks, the breach almost always occurs in between.
Again, according to Verizon's DBIR, even the top-performing organizations only address 30-40% of known-exploited vulnerabilities within the first week after detection: a rate that has barely changed despite years of consistent investment.
So, instructing teams to patch faster does not alter the underlying reality, and it resembles asking a freighter to stop on a dime.
The Bottleneck Shifted. So Must the Approach.
For two decades, vulnerability management relied on a straightforward set of assumptions:
Identify the flaws,
Rate them by severity,
Patch the most critical ones first.
When a few dozen critical vulnerabilities appeared each quarter, CVSS triage worked. Unfortunately, it is ineffective against hundreds or thousands of disclosures daily.
Returning to Verizon's DBIR once more, the median organization needed to patch 16 known-exploited vulnerabilities in 2025, up from 11 the previous year, an increase of nearly 50%.
And that was before AI-discovered flaws began flooding the catalog.
Severity scores, meanwhile, do not indicate whether a flaw is accessible in your environment, whether your controls will block it, or whether it connects to anything significant. A severity list where everything is a "9" or "10" essentially prioritizes nothing.
So the relevant question shifts from "what is vulnerable?" to "what is actually exploitable against us right now, and would our defenses catch it if someone tried?"
This is precisely the question that Breach and Attack Simulation (BAS) was designed to address.
Why BAS Becomes the Cornerstone Against AI-Powered Attacks
BAS applies real-world adversary techniques, the TTPs behind the latest headline campaign, and safely runs them against your live prevention and detection stack. Not a scan. Not a theoretical mapping. An actual exercise that demonstrates what your tools will actually block, what they will detect, and what will get through.
In a world overwhelmed by disclosures, this accomplishes three things that vulnerability management alone cannot. BAS:
Separates the theoretical from the real. A flaw that your WAF, IPS, and EDR already neutralize is a very different issue from one that enters directly. BAS shows which is which, so teams stop treating every CVE as a critical emergency.
Validates the controls you have already invested in. Most enterprises run anywhere from ten to seventy security tools with countless overlapping policies; BAS measures whether they function as configured and reveals the residual risks hiding in the gaps.
Buys time to patch safely. When you can demonstrate that a critical asset is already protected by hardened controls, the patch can proceed through normal change control instead of an emergency rollout. When it is not covered, you know to mitigate first.
This benefit is starting to appear in budgets: field reports increasingly indicate that CISOs are reserving dedicated spending for BAS that was not a separate line item a year ago.
This is the shift that Gartner now calls Adversarial Exposure Validation: combining security effectiveness ("Are my controls working?") with business context ("Which assets matter most, and what is truly reachable?") to prioritize based on your organization's reality rather than hypothetical raw scores.
Paired with autonomous penetration testing, which verifies whether an attacker can chain exposures from their initial foothold to your organization's crown jewels, BAS completes the picture.
One side asks, "Wait, can they breach us?" The other asks, "But would we catch it?"
Running together, BAS and autonomous pentesting replace guesswork with evidence.
BAS Has to Run Autonomously at Machine Speed Too
There is a catch.
If adversaries are operating autonomously, a validation cycle that takes a human a week to complete is obsolete upon arrival. Machine-speed attacks require machine-speed defenses, and the only thing fast enough to counter autonomous offense is autonomous defense.
The honest objection to directing raw generative AI at this is safety. As Picus CTO Volkan Erturk has warned, a model told to invent an exploit might return a live malware sample, or hallucinate techniques a group never uses. You do not want unvetted binaries detonating in production, or defenses built against attacks that do not, or cannot, exist.
Picus' solution is to put the model in charge of coordination, not creation.
Instead of asking AI to write payloads, Picus' agentic BAS matches a fresh threat report against a curated, pre-vetted library of safe, ready-made test building blocks. A security team names a threat, and a multi-agent system takes it from there: one agent identifies the threat and builds a research plan, others gather and validate the intelligence from multiple sources, and a builder agent maps the adversarial TTPs into attack chains ready for simulation.
The output is an accurate, ready-to-run simulation, assembled in minutes.
This shortens the loop. A CISA alert or a forwarded headline becomes a scoped test, a posture score, prioritized mitigations, and an executive report, often in minutes, with humans reviewing exceptions rather than driving, and slowing down, every step.
This Is What the Picus Platform Is Built For
Patching is still essential, but where AI discovers flaws by the thousands and weaponizes them in hours, patching alone cannot be your entire strategy. If the offense is autonomous, the defense must operate at least at the same speed, and that is exactly what Picus was designed to do.
What scales with the threat is validation: confirming what your controls will actually stop, proving what is exploitable, and spending remediation time and talent only where it will change the outcome. AI-powered, agentic BAS is one of the core pillars of the Picus Platform, continuously testing whether your defenses block and detect what matters without waiting on a human to start the process or move to the next cycle. And when a gap is uncovered, the platform points to the vendor-specific mitigation needed, and does not just create another ticket on the pile, then re-validates to confirm that the gap has actually been closed.
The need to say, on the spot, whether a fresh headline puts the business at risk is not going away anytime soon. The Picus Platform gives security teams that answer before anyone asks.
Find out if the next headline puts you at risk, before it drops. Request a demo.
Note: This article was written by Sıla Özeren Hacıoğlu, Security Research Engineer at Picus Security.