CVE-2026-72777 is a server-side request forgery flaw in Next AI Draw.io through version 0.4.16. The parse-url endpoint checks hostnames as strings without resolving DNS, so unauthenticated users can reach internal services. The record scores it 8.6, High. No fixed version is confirmed, so check the project.
What's affected
The record says Next AI Draw.io through 0.4.16 is affected. The project issue that reports the flaw lists affected versions as 0.4.16 and earlier, and shows the patched version as pending. We found no fixed release in the sources we read, so do not assume that a newer version is safe. Check the project's releases and issue tracker for a fix before relying on one.
The vulnerable route is POST /api/parse-url. The issue explains that the private address check compares the hostname with a list of known string patterns and never resolves it. A hostname that looks harmless but resolves to an internal address passes the check. The record says unauthenticated attackers can reach arbitrary internal HTTP services and read responses, including cloud metadata.
The weakness class is CWE-918, server-side request forgery. The issue also lists CWE-350, reliance on reverse DNS resolution for a security-critical action.
How to check if you're exposed
Find out whether you run this application and where it is exposed.
npm ls next-ai-draw-io
grep '"version"' package.json
Run these in the project directory of your deployment. If you use a container image or a hosted build, check the tag or commit against the project's releases. If the version is 0.4.16 or earlier and the app is reachable from the internet, treat it as exposed.
Then consider what the host can reach. Server-side request forgery is only as damaging as the network around the server. A host in a cloud account that can read an instance metadata service, or one that sits next to internal admin tools, has more to lose. Your network team can confirm what the application host is able to connect to.
What to do now
- Check the project releases and the issue for a fixed version, and upgrade when one is confirmed.
- Until then, do not expose the application to the internet. Put it behind authentication or a private network.
- Block the application host from reaching internal address ranges, link-local addresses and the cloud metadata service at the network level.
- If your cloud provider offers a metadata service mode that needs session tokens, enable it for the host.
- Disable or remove the parse-url route if you do not use it and can change your deployment.
- Review outbound request logs from the host for requests to internal addresses.
Why it matters
Checking a hostname as text is not the same as checking where it leads. A name can resolve to a public address when the check runs and to an internal one when the request is made, or simply point at an internal address from the start. A safe check resolves the name, tests the resulting address, and connects to that same address.
On severity, the record scores it 8.6 High under CVSS 3.1, with a changed scope. VulnCheck lists it as high at 7.7 under CVSS 4.0. The two are close in meaning. The key difference from other items this month is the lack of a confirmed fix. Where no patch exists, network controls and removal of exposure are your main defenses. No source reports exploitation in the wild, and the CVE is not in the CISA KEV catalog.
Try it yourself
Click through it at your own pace, or open it full screen.
Quick check
Is there a fixed version of Next AI Draw.io?
We could not confirm one. The project issue lists the patched version as pending. Check the project for updates.
Why is cloud metadata mentioned?
The record says attackers can reach internal HTTP services and exfiltrate responses, including cloud metadata, which can hold credentials.
Sources
Last reviewed: 18 October 2026.
