- Vercel says Next.js advisories are rising because AI makes old bugs cheaper to find. It reports 70% of this year's 51 flaws were in code at least a year old.
- The claim is Vercel's own. It describes a bounty, an outside audit, AI review of every pull request, and monthly security releases announced in advance.
- Vercel says AI also makes old bugs cheap to exploit; the implication is that unpatched apps are more exposed. Teams should treat upgrade speed as the control.
A longer advisory list can mean better searching
Next.js has published more security advisories in recent years. Vercel, the company behind the framework, says the main reason is that AI has made bugs easier to find. In a post on the Next.js blog, it said 51 vulnerabilities were discovered this year.
Vercel also gives the age of those flaws. It says 70% of the 51 were in code at least a year old. Within that group, 43% of the total came from code three years old or more. In other words, most of this year's findings were not fresh mistakes. They were long-standing ones.
Here is the idea worth holding. An advisory count measures how cheap it is to look, as well as how much is broken. But cheap tools work for both sides. Vercel says AI has made finding and exploiting old bugs cheap. In its words, edge cases have become "risks that are now worth fixing".
Vercel's own account is that the recent rise reflects bugs in older code that went unnoticed for years. That is the company's view of its own project. The post does not offer independent measurement, and it does not say how many bugs remain.
Vercel says AI also makes old bugs cheap to exploit; the implication is that unpatched apps are more exposed. That is WebPulse's inference from Vercel's claim, not a figure the post gives.
What WebPulse data shows about scale
WebPulse's NVD profile, refreshed on November 18, 2026, lists 63 CVEs in total for Next.js. It counts 39 of them in the last 12 months. Those figures come from a different method than Vercel's count of 51, so the two should not be set against each other.
They give a rough sense of the footprint. WebPulse's September 2026 scan of the Tranco top 10,000 domains had 7,064 responding sites. A platform was detected on 2,491 of them. Next.js was detected on 791, or 31.8% of those. In that sample, many large sites face this upgrade cadence.
How Vercel says it is finding the old bugs
Vercel describes four sources. The first is a bug bounty program. Its stated goal is to make reporting a bug privately more profitable than exploiting it.
The second is an internal bot built with Eve. Vercel says it finds vulnerabilities, triages reports, builds reproductions, affirms severity and checks whether fixes are comprehensive. The third is a manual audit by the independent firm Trail of Bits, which Vercel says found issues its other sources had missed.
The fourth guards new code. Vercel says it now runs an AI security reviewer, part of Vercel Agent, on each pull request to Next.js. The reviewer reads every change as an attacker would and tries to break it. Vercel says it has already caught issues in pull requests that had passed code review and automated tests.
Where the risk sits: the seams and the parts
Vercel names two structural weak points. The first is the boundary between Next.js and the hosting layers around it, such as proxy routing, content delivery networks and JavaScript runtimes. Vercel says disagreements between these layers can allow a bypass. Its answer is Adapters, a defined contract telling hosting providers how Next.js apps should be served.
The second is third-party code. Vercel says many of this year's vulnerabilities came from image libraries it depends on, not from Next.js itself. Those libraries handle untrusted input and are large and hard to review. Vercel says it is moving to memory-safe image codecs, which are built to rule out a common class of bugs. It says AI helped it adopt them faster. It has not yet published results.
Vercel also says it is confident the yearly advisory count will change direction as old bugs are cleared. That is a company expectation, not a measured result.
The upgrade is the control you own
Vercel now ships security releases roughly monthly and announces them in advance. The aim is that customers can plan, and hosting providers can add protections before details go public.
Starting with Next.js 16.4, the development and build commands warn when the running version has a known vulnerability. Vercel also describes an upgrade command. It checks your version against the advisories and chooses the newest fixed release without forcing a jump to a new major version. It then passes the work to a coding agent along with migration guides. An agent's changes still need your own tests to confirm the app works.
Questions to put to your team
Ask which Next.js version each production app runs, and how far it trails the latest security release. Ask who owns the monthly release, and whether a patch can ship within days.
If a provider hosts your apps, ask whether it receives advance notice and what protections it adds. Ask whether your tests are strong enough to trust an agent-run upgrade. Ask which features, such as image handling, touch untrusted input.
A long advisory list may mean someone is looking harder. What protects you is how quickly your organisation applies the fixes.
Produced by the WebPulse Newsroom with AI assistance from the original reporting credited below, and checked against that source by our editorial review. How we use AI.
Original reporting: Next.js (Vercel).





