Datasette 1.0a39 and 0.65.4 security releases
Simon Willison’s Weblog Simon Willison
Datasette shipped two security fixes for its alpha and stable lines. If you run it on the public web, especially with mixed public and private tables, update now.
Based on reporting by Simon Willison’s Weblog, Simon Willison — read the original for the full story.
Summary, retelling and take written by AI under human oversight; images are AI-generated illustrations. How we work · Report an error
Simon Willison has shipped security patch releases for both branches of Datasette: 1.0a39 for the current alpha series and 0.65.4 for the stable 0.65.x line. The warning is aimed squarely at anyone running a Datasette instance on the public web, and especially at setups that mix public and private tables.
The fixes came after issues were reported by Sevban Dönmez. Willison says Alex Garcia and he then ran an extensive audit of Datasette with Claude Fable 5.1, GPT-5.6, and GPT-6 Astra, spending almost a week on the follow-up work and on reviewing the resulting fixes. Some of the bugs were subtle enough to need that kind of pass.
The process matters as much as the patches. Willison says security audits by frontier models will now become part of Datasette development going forward.
Garcia also suggested splitting the work in a way that made the review tighter. They used a shared private repository, divided tasks so one person wrote automated tests to expose an issue and the other implemented the fix, and made sure two humans looked at each problem as well as the coding agents.
My take — AI-written commentary, not fact-checked reporting
This is the sane version of AI-assisted security work: use the models, but don’t let them pretend they’re the adults in the room. The interesting bit is the split-review setup, not the model names on the tin. Anyone shipping software with public and private data mixed together should probably be stealing that workflow by lunchtime.
Read more about this at: Simon Willison’s Weblog
