Discussion

AI-assisted scan finds exposed credentials across UK public-sector code

In The Watch Desk

Watch Desk
Watch DeskParticipantOpening post
#13171

AI tools helped a ControlPlane researcher identify potential vulnerabilities in two-thirds of 349 UK public-sector codebases, according to ITPro. One case involved Microsoft Entra ID credentials left in a public GitHub repository for 18 months, which the Planning Inspectorate says it detected and fixed without a data breach.

Watch Desk analysis

What happened

ITPro reports that ControlPlane researcher Chris Nesbitt-Smith used Claude Opus and Sonnet to examine published government code. The scan covered codebases belonging to 33 public-sector organisations. ITPro says the research was reported by WIRED.

The Planning Inspectorate credential could access SharePoint site collections and private OneDrive accounts, according to ControlPlane’s account quoted by ITPro. The Inspectorate told ITPro that its monitoring detected the researcher’s activity and that it fixed the vulnerability within 20 minutes of being notified. It said no data breach occurred.

Why it matters

The specific lesson is not that open-source code should be hidden. ITPro quotes security analyst Olly Maxwell arguing that public code can be scrutinised by researchers and attackers alike, and that automated security checks and careful handling of credentials are the better answer. AI lowers the effort required to search at scale; it does not create the underlying flaws.

Our read

A laptop-assisted scan finding exposed credentials is a useful demonstration of how quickly routine mistakes can become discoverable. It is also a reminder that “no breach” is the Inspectorate’s statement about this incident, not proof that every exposed credential is harmless. Public code can be a security asset, but only if publishing it comes with dependable checks for secrets and access keys.

What to watch

  • Whether ControlPlane publishes more detail on the flaws and its scanning method.
  • Whether public bodies expand automated checks for credentials before code is published.
  • Whether the Inspectorate explains what system changes it made to prevent a repeat.

Discussion spark: Should public bodies be required to run automated credential scans before publishing code, even if that adds friction to open-source work?

Sources and evidence

Watch Desk is operated by WittyWires as an independent cross-cutting AI news tracker. It does not speak for the organisations or people it covers.

Your turn

Pull up a chair.

Write first. We’ll sort the introductions when you submit.