How urgent is the Datasette upgrade?

Datasette's maintainers say you should upgrade now if you run an instance on the public internet, and in particular if you use a Datasette authentication plugin to protect private data. A purely local or fully public instance with no private tables is at much less risk, because the fixed bugs mostly involve permission checks leaking data a visitor should not reach.

Is there a CVE or a public exploit for these bugs?

Neither the Datasette release post nor Simon Willison's write-up lists CVE identifiers. Alex Garcia and Simon Willison say they are holding back some of the automated tests from the public repo so that operators have more time to upgrade before the exact reproduction steps are visible to everyone.

Did the coding agents find these bugs on their own?

Not alone. Claude Fable 5.1, GPT-5.6 Sol and GPT-6 Astra drove the audit over several rounds, but Alex Garcia and Simon Willison split the work so that one wrote an automated test proving an issue and the other implemented the fix. That means two humans reviewed each issue in addition to the agents.

Should I move from Datasette 0.65.x to the 1.0 alpha?

You do not have to. Datasette shipped a patched version on both lines the same day — 0.65.4 for the stable series and 1.0a39 for the current alpha series — so staying on 0.65.x is a supported way to get the fixes. Pick the line you already run and take its patch release.

Datasette · 2026-09-11 · major

Datasette 1.0a39 and 0.65.4 — security fixes from a coding-agent audit

Datasette 1.0a39 and 0.65.4 are security releases that close permission and escaping holes across table, view and search endpoints. Simon Willison and Alex Garcia found them in the project's first coding-agent security audit.

Datasette project card from datasette.io

Datasette's first thorough coding-agent security audit turned up subtle permission bugs, now patched in two releases.

Quick facts

Project	Datasette
Versions	1.0a39 (alpha), 0.65.4 (stable)
License	Apache-2.0
Who should upgrade	Public-internet instances, especially with private tables
Reported by	Sevban Dönmez
Audited with	Claude Fable 5.1, GPT-5.6 Sol, GPT-6 Astra

What is it?

Datasette 1.0a39 and 0.65.4 fix bugs that could let a visitor read data they should not see. Most sit in permission checks: table and view names are now matched case-insensitively the way SQLite treats them, full-text search index tables require permission on the source table, and the sqlite_stat1 to sqlite_stat4 tables are denied by default. Datasette is an open source tool for exploring and publishing data held in SQLite.

How does it work?

Sevban Dönmez sent in several AI-assisted vulnerability reports, which prompted Alex Garcia and Simon Willison to audit the codebase with Claude Fable 5.1, GPT-5.6 Sol and GPT-6 Astra. Several rounds of auditing turned up more problems. One person wrote an automated test that showed an issue and the other wrote the fix, so two humans reviewed every item alongside the agents. Some of those tests are being held back from the public repo.

Why does it matter?

Anyone running Datasette on the public internet should upgrade now, and that goes double for instances where an authentication plugin protects private tables sitting next to public ones. The wider point is about method: this was the project's first thorough coding-agent security audit, and it surfaced real, subtle bugs that had survived years of ordinary review.

Who is it for?

people self-hosting Datasette

Frequently asked questions

How urgent is the Datasette upgrade?: Datasette's maintainers say you should upgrade now if you run an instance on the public internet, and in particular if you use a Datasette authentication plugin to protect private data. A purely local or fully public instance with no private tables is at much less risk, because the fixed bugs mostly involve permission checks leaking data a visitor should not reach.
Is there a CVE or a public exploit for these bugs?: Neither the Datasette release post nor Simon Willison's write-up lists CVE identifiers. Alex Garcia and Simon Willison say they are holding back some of the automated tests from the public repo so that operators have more time to upgrade before the exact reproduction steps are visible to everyone.
Did the coding agents find these bugs on their own?: Not alone. Claude Fable 5.1, GPT-5.6 Sol and GPT-6 Astra drove the audit over several rounds, but Alex Garcia and Simon Willison split the work so that one wrote an automated test proving an issue and the other implemented the fix. That means two humans reviewed each issue in addition to the agents.
Should I move from Datasette 0.65.x to the 1.0 alpha?: You do not have to. Datasette shipped a patched version on both lines the same day — 0.65.4 for the stable series and 1.0a39 for the current alpha series — so staying on 0.65.x is a supported way to get the fixes. Pick the line you already run and take its patch release.

Try it

pip install -U datasette