AI is supercharging vulnerability discovery. What that means for state, local governments amid wave of outages
As artificial intelligence tools accelerate the discovery of software vulnerabilities, state and local governments should prepare for more frequent system disruptions and maintenance-related downtime, cybersecurity experts warned Wednesday.
Experts from the Center for Internet Security said governments should begin treating outages as a “new normal” as the pace of vulnerability discoveries increases. A wave of disruptions over the past week affecting Microsoft, Workday, JPMorgan Chase, Apple, United Airlines and the Federal Aviation Administration were attributed to IT administration issues rather than cyberattacks. But experts said the availability of critical systems is itself a cybersecurity concern, underscoring the need for governments to strengthen basic security practices and prepare for disruptions.
A number of these outages impacted state and local governments, said Randy Rose, vice president of security operations and intelligence at the Center for Internet Security. The Microsoft outage caused many government organizations to lose access to Outlook for email. In some cases, they lost access to SharePoint or other file-sharing services, as well as Teams, another Microsoft communications service.
Rose said this highlighted the need for governments to begin thinking about a backup communications plan in the event of outages. But more than that, governments also need to contend with the potential increase in outages. The spike in vulnerability disclosures isn’t because software suddenly got worse, Rose said. Instead, he attributed this sharp uptick to the fact that now, more people — and now more advanced AI tools — are looking for vulnerabilities all at once.
To illustrate that point, Rose said vulnerability disclosures rose from about 4,274 in October 2025 to 17,060 in October 2026, adding that about 90% of disclosed vulnerabilities are not practically exploitable, and therefore the increase is not a reason to panic outright.
Big tech vendors and software makers, such as Microsoft, Adobe and Oracle, along with thousands of smaller vendors, routinely find and fix bugs in their own products. They also hire internal teams or outside firms to continuously scan their networks. When they do, they publish advisories and register them in public databases as “CVEs,” or common vulnerabilities and exposures, Rose said.
Independent researchers are also currently getting paid to find flaws through bug bounty programs. As their tools improve, including AI-assisted code analysis and fuzzing, “they’re uncovering more problems in widely used software and cloud services,” Rose continued.
Additionally, automated scanning tools, now made more powerful by AI, are becoming more widely available. Rose said that though organizations have run automated scanners to look for security issues in code and systems for years, those tools are now being upgraded with AI, which can spot patterns faster and generate proof-of-concept attacks more easily.
For state and local governments, most of these vulnerabilities are being found upstream, in commercial and open-source products they depend on — not necessarily in systems they wrote themselves, Rose added. But every new disclosure still shows up as one more patch to track, test and install, and more pressure on already thin security teams.
“Ultimately we’re kind of living in a new normal. We know that AI is rapidly accelerating the pace at which vulnerabilities are disclosed, so we have to get used to seeing those numbers increase significantly,” Rose said. “What it is is an opportunity for organizations to look at what trusted best practices, trusted guidance — whether it’s from vendors directly, from organizations like ours or whatever their information standard is — and really get those basics right. You can reduce a significant amount of risk, even in a high‑vulnerability environment, by putting a relatively small set of controls in place comparatively.”
The need to “get those basics right” is made more urgent, Rose said, when considering that some threat actors did attempt to “take advantage of the chaos” over the past week’s wave of outages.
Rose said CIS saw reports of some social engineering impacts and threat actors standing up typo-squatted domains, which deploy malware to those who open infected links. CIS’s multi-dimensional threats team also looked into the physical impacts to transportation, such as the over 10,000 flights that were canceled, and how that influenced perception of the problem.
Given the scale of the events, Rose said it was important to acknowledge the flurry of conspiracy around them, with some looking for links. In a briefing with state, local, tribal and territorial cyber leaders, he said that the governments were briefed on how information spreading online, often false information, also amplified the events.
“That same team covered some of the narratives that we’re seeing around the global blackout. This is all stuff that’s really kind of in that information operation space. Sometimes it’s foreign adversary amplified, but not always. Sometimes just internet conspiracy theories that take off. So we were really starting to see a lot of those unsubstantiated online narratives taking off, really around this Nov. 17 global outage,” Rose continued.
“Obviously, that didn’t happen, but those were things that you know lined up with the FEMA emergency test, so people were jumping on that, like, ‘Oh, it’s too coincidental,’ and people really jumped on that. So a lot of confirmation bias happening there — and then at the very end, we just wrapped up with a message, really a message of calm… We discussed how outage clusters do happen. A year ago, we saw multiple cloud providers in a very short period of time, about 30 days, all had major outages right around the same time. And sometimes when one outage happens, people tend to increase their reporting of outages, so sometimes bystander services get blamed. We also talked about outages being very normal during upgrade windows, and the IT administrator knows this.”
Phyllis Lee, vice president of security best practices content development at CIS, said that in light of these events and the turbocharged discovery of vulnerabilities, state, local, tribal and territorial governments should focus on resilience, not on specific threat actors.
She advised organizations to patch and upgrade carefully, map their third-party dependencies, monitor providers for outages and patches, and plan how they’ll notify staff and residents if services fail.
Particularly, she pointed out that because governments can be very reliant on third-party providers, it’s crucial to clearly understand which services those vendors support, especially if they are critical resident services, so they can assess the impact if a provider goes down and come up with backups.
“Are you monitoring those third-party providers to see, you know, are they supplying patches? Are there outages? And so, then how are you going to deal with that in case something goes wrong? How is it that you can notify your employees? How is it that you can notify your own customers?” Lee said, in describing the questions governments should be asking themselves. “Do you have redundancy built into your system? Do you have workarounds? … [And] through the power of purchasing, [demand] that our service providers … provide software that is secure by design.”