Futureweb

Software

Python's October releases distrust a named CA

Python shipped 3.15.2 and 3.14.10 with installers, plus source updates on three older lines. One change distrusts compromised eMudhra intermediates. The platform store still leads.

The short version

No. Hugo van Kemenade's October 26, 2026 post covers 3.15.2, 3.14.10, 3.13.18, 3.12.16, and 3.11.18. Binary installers are for 3.15.2 and 3.14.10. The other three are source-only. All five distrust compromised eMudhra intermediate keys in the ssl module, as a last resort behind the platform trust store. The changelog dates the 3.15.2 final to October 25.

Five versions, two installer lines

The Python blog post of October 26, 2026, by Hugo van Kemenade, announces 3.15.2, 3.14.10, 3.13.18, 3.12.16, and 3.11.18 together. Installers are named for the first two. The older three are source releases. The changelog's date for the 3.15.2 final is October 25, a day before the blog post. Discourse carries the same text as the blog. A memory-safety note is the standing reminder that a version number is the fact. The document named in the sources is the record this page will keep.Read the version in the note.

The denylist is a last resort

CVE-2026-115557 is the identifier on the ssl change. The post says Python will distrust the compromised eMudhra intermediate keys, matching on the public key, and that this is a last resort because the operating system's trust store is the check that matters first. The earlier brief is the one that records eMudhra leaving the public TLS business. This page is the runtime that added a backup refusal. The document named in the sources is the record this page will keep. The document named in the sources is the record this page will keep.The exit was the earlier brief.

Two other names, and no recipe

CVE-2026-114583 is the tarfile filter that rejects links replacing the destination. The cookie change rejects a negative version number. Those are the changelog's names. This page does not expand them into a method. An advisory is the vendor's story when the vendor is the one writing it, and here the vendor is the Python project. The document named in the sources is the record this page will keep. The document named in the sources is the record this page will keep. The document named in the sources is the record this page will keep.The project post is the advisory.

Older lines already had the library bumps

The post says Expat 2.9.0, the Windows mimetypes known-files change, and header refolding were already in 3.15.1 and 3.14.9. Readers on those two lines should not treat every bullet in the roundup as new today. Readers on 3.13, 3.12, and 3.11 are being offered the source trees that carry the set. A CVE identifier is the label, not the patch. The document named in the sources is the record this page will keep. The document named in the sources is the record this page will keep.An identifier is the label.

Who should read which file

People who install from python.org care about 3.15.2 and 3.14.10. People who build from source care about the three older tags as well. A distribution that has not yet packaged the tags is not updated because this post exists. Asked directly, What was already in 3.15.1 and 3.14.9? The post points at Expat 2.9.0, a Windows mimetypes path fix, and header refolding as items those earlier releases already carried. They are not new in the October 26 post's sense of newly added to those two lines.

Node's October patch is a different runtime

An October release for Node was a different project, a different set of versions, and a different brief. Sharing a month does not share a changelog. Python's denylist does not patch a JavaScript runtime. Asked directly, Which versions have installers? 3.15.2 and 3.14.10. The post says 3.13.18, 3.12.16, and 3.11.18 are source-only. Asked directly, Is this the same story as eMudhra leaving the public TLS business? It is the next document. That brief is the company's exit. This one is Python refusing the intermediates. A release note about memory safety is the habit both pages share: read the version.October also had a different runtime.

What the Discourse copy adds

The forum post is the blog post again, in the place the core team discusses releases. It corroborates the text. It does not add a sixth version. A short kernel slice is unrelated work in the same season. Asked directly, What else do all five releases share? The changelog names CVE-2026-114583, a tarfile filter that rejects links which replace the destination, and a cookiejar change that rejects a cookie with a negative version. The mechanism beyond those names stays in the changelog. The document named in the sources is the record this page will keep.A kernel slice is not this runtime.

Platform store first

The post is explicit that distrust inside Python is not a substitute for the operating system doing its job. An administrator who only updates Python and leaves a broken platform store has done the last-resort half. Asked directly, What is CVE-2026-115557 in this release? The changelog's name for distrusting compromised eMudhra intermediate certificate-authority keys inside Python's ssl module. The post says the platform trust store remains the primary check, and the denylist matches a public key. The document named in the sources is the record this page will keep.

What this page will not expand

How a tar link is built, and how a cookie version is parsed, stay in the changelog and the commits. The news is which versions contain the named changes. The short answer this page will stand on is this: No. Hugo van Kemenade's October 26, 2026 post covers 3.15.2, 3.14.10, 3.13.18, 3.12.16, and 3.11.18. Binary installers are for 3.15.2 and 3.14.10. The other three are source-only. All five distrust compromised eMudhra intermediate keys in the ssl module, as a last resort behind the platform trust store. The changelog dates the 3.15.2 final to October 25.

Sources

The reports this brief is filing. Futureweb did not republish them.

  1. Python blog
  2. Python changelog
  3. Python Discourse

Questions

Which versions have installers?

3.15.2 and 3.14.10. The post says 3.13.18, 3.12.16, and 3.11.18 are source-only.

What is CVE-2026-115557 in this release?

The changelog's name for distrusting compromised eMudhra intermediate certificate-authority keys inside Python's ssl module. The post says the platform trust store remains the primary check, and the denylist matches a public key.

What else do all five releases share?

The changelog names CVE-2026-114583, a tarfile filter that rejects links which replace the destination, and a cookiejar change that rejects a cookie with a negative version. The mechanism beyond those names stays in the changelog.

What was already in 3.15.1 and 3.14.9?

The post points at Expat 2.9.0, a Windows mimetypes path fix, and header refolding as items those earlier releases already carried. They are not new in the October 26 post's sense of newly added to those two lines.

Is this the same story as eMudhra leaving the public TLS business?

It is the next document. That brief is the company's exit. This one is Python refusing the intermediates. A release note about memory safety is the habit both pages share: read the version.