Skip to content
Humanopia
Safety

New study measures how often chatbots make things up

Error rates have fallen, but not to zero, and they vary by topic.

Humanopia Editorial Team · Sep 13, 2026 · 1 min read

Prices did not change, but limits did. Heavy users may hit the new caps sooner, especially when they upload large files or run long tasks in the background.

We will update this story as more independent testing is published. If you have used the feature and want to share what you found, write to the editorial team.

The details

The announcement came with a short blog post, a technical document and a handful of demos. As usual, the demos show the best case. We read the documentation to see what the product does on an ordinary day.

  • No change to default privacy settings
  • Older versions stay available for now
  • Independent tests are still limited
  • Rolling out region by region

The bigger picture

Independent researchers were quick to test the claims. Early results broadly match what the company said, with the usual caveat: benchmarks measure narrow skills, and real work is messier than a test set.

Stay ahead without
living inside the feed.

Get our weekly newsletter with the most important AI news, practical guides and thought-provoking essays, curated for curious, forward-thinking people.

Weekly, not daily Curated and practical No spam, ever