Critical flaw in Taipy's web server config lets any website hijack your session, CVE says
A newly disclosed bug in the Python framework Taipy scores a near-maximum severity rating - but whether it's actually being exploited, or even patched, is still an open question.
A newly published vulnerability, CVE-2026-85183, has landed with a CVSS score of 9.3 - the kind of number that usually means “patch immediately”. It affects Taipy, an open-source Python framework popular with data scientists for wrapping their models and dashboards in a web front end without having to learn proper web development. Before anyone panics, though, it’s worth being precise about what’s actually confirmed here.
What the bug actually does
According to the NVD entry, Taipy sets up its socket.io server - the bit that handles live, two-way updates between a browser and the app - with a wildcard CORS origin (meaning any website, anywhere, is treated as an allowed origin) while also enabling credentials on those connections. Put simply, that combination tells a user’s browser it’s fine to send authenticated, credentialed WebSocket traffic to a Taipy app from literally any other web page the user happens to have open. That’s a textbook cross-origin misconfiguration: CORS wildcards and credentials are meant to be mutually exclusive precisely because combining them removes the safety barrier between sites.
The relevant code sits in Taipy’s server.py, part of the project’s GUI module, which handles the socket.io setup for Taipy-built apps. NVD’s summary doesn’t spell out step-by-step how an attack would run beyond that setup, so the specifics of what an attacker could pull off in practice - reading data, triggering actions on someone’s behalf - aren’t laid out in the disclosure text itself, only that the credentialed cross-origin connection is possible.
So who is actually at risk
This isn’t a bug in a website you visit - it’s a bug in a piece of infrastructure. Taipy is a library that developers embed inside their own applications; Avaiga’s repository describes it as a tool for turning data and AI projects into “production-ready web applications”. So the people exposed aren’t Taipy’s own maintainers or a single hosted service, but anyone running a Taipy-based app that a real user might log into - internal dashboards, data tools, AI demos - with the vulnerable server configuration switched on.
Crucially, several things that would tell us how serious this is in practice are not confirmed in the disclosure: whether a fixed version has been released, whether the flaw is being actively exploited anywhere, and how many live installations are actually running the vulnerable code. None of that is stated in the NVD record or the linked GitHub pages, so treat those as open questions rather than settled facts.
What to do about it
If you build or run anything with Taipy, the sensible move is to check your version against Taipy’s own security advisories and release notes for a fix, and in the meantime review your own CORS configuration rather than assuming the framework’s defaults are safe. Ordinary end users of Taipy-built apps have no direct action to take - this is squarely a developer and system-administrator problem, not something to patch on your own laptop.
A 9.3 severity score reflects how bad the flaw could be if triggered, not proof that it has been. Until there’s confirmation of a patch and evidence of real-world exploitation, this is one for the people who deploy Taipy to check quietly and fix - not a reason for anyone else to lose sleep.