Skip to content
Humanopia
Safety

Researchers find new jailbreak that works across models

A single technique bypassed safety filters on several assistants. Labs say fixes are rolling out.

Humanopia Editorial Team · Oct 6, 2026 · 1 min read

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.

Regulators have not commented yet. In the EU, new obligations for general-purpose models mean that documentation like this is now expected rather than optional.

How we reported this

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.

  • Available first to paid plans
  • Independent tests are still limited
  • Admin controls for business accounts
  • Usage limits apply to long tasks

The bigger picture

Not everyone is convinced. Some analysts say the update is mostly about catching up with rivals rather than moving the field forward. Others point to small but useful changes that the headlines missed.

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