Like Christmas decorations at Costco, security prediction season kicks off earlier and earlier every year. I could probably write half the posts for you right now. Yes, AI-driven attacks are going to get scarier and yes, an acronym no one has heard of in January will be on every booth banner at RSA by April. I want to focus on what these forthcoming posts are going to miss (along with most of what I've seen written for Cybersecurity Awareness Month). It's where the real damage starts: the vulnerability findings a team already knows about…but just hasn't gotten around to yet.

I've spent more than twenty years on both ends of the findings report, but the backlogs I hear about from security teams today are deeper and represent a broader attack surface than we've ever seen. It's truly unprecedented. Here's what I think we'll see in 2027, and what I'd recommend doing about it.

Once-a-year pentests run out of road

The annual or routine pentest is still common, with companies running them and fixing whichever findings seem most critical. The rest just sit until next year's test finds them again. Letting findings linger in your backlog has always been an expensive way to run things, but now it's also a lot more dangerous because a moderately capable attacker with a moderately capable AI model can achieve a working exploit of new vulnerabilities within a day.

Fellow chess players out there would call it a "tempo problem." The attacker already gets to move whenever they like, and a 12-month testing cycle gives them free move after free move before the defense responds. Something's got to give, and I think more corporate boards (and cyber insurers) will ask how often environments are getting tested as well as how the organization is prioritizing the ever-growing backlog of exposures. Answering "once, back in Q2" isn't sufficient. Testing needs to closely track changes in the environment (big releases, a cloud migration, etc.), versus a calendar date.

Fixed, or "fixed"? Someone's checking

Automated remediation using AI is going to be on the rise. AI may be able to facilitate so many patches in 2027 that, in many shops, they'll merge with barely a human glance. I understand the appeal. Backlogs are thousands deep and teams are a handful of folks, if even that. But marking a ticket "resolved" (whether by machine or one of us humans) only tells you a change was made. Did the exposure go away? Will it return? Did the fix cause other exposures? Those are separate questions.

Back in my testing days we'd routinely retest "closed" items and find them still wide open or exploitable via a slightly altered technique. A patch passed the check the scanner ran, perfectly. It just didn't pass anything beyond the scan it was built to pass. Every finding comes down to two questions. One, who's fixing it, and two, is it really fixed? We are going to see a lot more questions being asked about the second one next year.

CVSS loses its grip on the queue

Severity alone is misleading (it's always been misleading). But NIST's move this spring to stop enriching most CVEs forced the issue for a lot of teams, who found out their SLAs and ticket routing depended on a score field that might never get filled in. So in 2027 I anticipate programs pulling ahead will be those finding a way to rank findings by whether they can be realistically exploited, how much the asset matters, what controls tie into it, and what the cost would be, to the business, if things went south. Context is king. And that context comes from your own business data and asset inventory, not a universal scoring system. It's where it lived already, but NIST's change should expedite more teams to think about this correctly.

Focus on ROI

I'm sure you have more security tools than you did five years ago. But how well can you show your risk went down because of it? Every scanner and pentest yields findings in its own format with its own severity scale, and then it gets stitched together in a spreadsheet. Spreadsheets often become the duct tape within any business process, but present significant risk. They aren't centralized and can be limited in the amount of detail needed to remediate the findings. Tracking that remediation also is incredibly manual.

If the spreadsheet owner is also one of your better analysts, you have another challenge. When every finding shows up looking equally urgent, remediation slows down and the people doing the work burn out. Fixing that is where I think the 2027 cybersecurity budget conversation will go. We'll continue to move away from what to buy that can identify more exposures and toward connecting what's already been bought. Removing silos and gaining visibility on return on investment (ROI) of security tooling is going to be a focus. CFOs are increasingly asking whether the last few years of security spending made the company measurably safer. Answering takes one view of exposure that pulls in findings from every source, while collapsing duplicates, tying each to an asset and owner, and following it from discovery to verified fix.

Finding counts get retired (as a scorecard)

Finding counts will never decrease and you can close a hundred low-severity items in a quarter and leave your exposure exactly where it was. However, a single fix on an internet-facing system with an active exploit can do more for your security posture than the rest of the quarter combined. When attackers are moving from discovery to exploitation as quickly as they now can, trying to close the most vulnerabilities is a losing game. Quantity doesn't necessarily mean improvement, and can be a dangerous red herring. I expect more security leaders to report on whether their most consequential risks went down, and how they can prove it.

Human testers get more valuable

You may disagree with me on this one, given how much of the industry is betting on AI. And I'll start by saying that AI is genuinely good at the wide, repetitive part of the work, like crawling big attack surfaces, or flagging the obvious stuff faster than we mortals can. It's also continuing to be impressive in finding new exploits at a much more rapid pace. But it still needs a human expert to architect and guide its testing. I think this is how things will play out in 2027: smart teams let AI handle the breadth and put their human testers on the depth and orchestration, which is roughly how good pentest shops already run, just with better tools underneath.

Two things you can do before January

If you want something more concrete than predictions, try these two before winter. First, pull up the oldest open critical finding in your environment and trace why it's still open. My bet is on it being an ownership question no one settled. Second, have someone retest a handful of recent remediations (preferably someone who didn't do the fixing) and see how many hold up.

Everything in your backlog has already been paid for and written up. Working it down, starting with the findings that can actually hurt your company, is how you pick the fights worth battling in 2027. It's also a lot cheaper than learning from an incident report which ones you should have picked.

About the author: Dan DeCloss is the founder and Chief Customer Officer at PlexTrac, the offensive security validation platform that's part of Brinqa. Dan started his career in the Department of Defense and then moved on to the private sector where he worked for various companies including Telos, Veracode, Mayo Clinic, and Anthem. Dan's background is in application security and penetration testing, involving hacking networks, websites, and mobile applications for clients. Prior to founding PlexTrac, Dan was the director of cybersecurity for Scentsy.

Dan DeCloss — Founder of PlexTrac https://blogger.googleusercontent.com/img/b/R29vZ2xl/AVvXsEiBem7s4I_LvTeBmWivAqOnaLWLB8cKfXw-7NiflOio7UNzyrSnXHvKFrIpKeZHpe6dCJ1hC94s-CGFULfTjLu-QGTTotxSRANNEj58jIRKY7aMSqaS1GJijPc-HrPDvhntXV4ommWPayFlnrDJkmATn7hyhu7BG2RF8MJ6U-x0jzZA0VITYyopQpvdnc0/s1700-e365/dan.png
Found this article interesting? This article is a contributed piece from one of our valued partners. Follow us on Twitter and LinkedIn to read more exclusive content we post.