
ControlPlane AI Security Team Responsibly Discloses Companies House's Emergency Authentication Code Vulnerability

A ControlPlane security research note. Reported and resolved through the Companies House vulnerability disclosure programme. Published with coordinated timing after remediation.
Summary
ControlPlane's AI security research team identified an authorisation gap in the Companies House Emergency Authentication Code (EAC) workflow. The workflow lets company officers regain access to their filing account when their normal credentials are lost; on approval, Companies House posts a paper letter containing the authentication code to the officer's registered address. The application did not verify that the person modifying an in-progress request was the one who created it, allowing an authenticated user to alter another user's request before submission.
Once reported, Companies House confirmed the issue and remediated it. Severity was assessed as High, CVSS 8.1 (CVSS:3.0/AV:N/AC:L/PR:L/UI:N/S:U/C:H/I:H/A:N), classified under CWE-862 (Missing Authorisation / Broken Object Level Authorisation).
Background
Emergency Authentication Codes are a fail-safe for company officers, and approval results in a physical letter being sent to a registered address, so the integrity of a request (i.e. who it is for, and where the code will go) matters. Registration for the service is free and immediate, meaning the population of authenticated users effectively comprises the public. The application decides which requests a given user is allowed to act on.
The finding
The workflow consists of two state-changing steps: selecting the officer the code applies to, and confirming and submitting the request. Each step referenced the request by an identifier in the URL.
The application checked the state of the request, for example, whether it had already been submitted, but didn’t check ownership: that the request belonged to the authenticated session making the change.
As a result, an authenticated user who held another user's request identifier could change the selected officer on that request from their own session, and the change persisted server-side on the original request.
Root cause
The application retrieved the caller's identity for display but never compared it against the owner recorded on the request. A business-state check had been mistaken for an authorisation boundary. State and authorisation answer different questions; only an ownership check answers "is this caller allowed to act on this object?", the most prevalent category of OWASP web risk, A01:2025 Broken Access Control (opens in a new tab).
Potential impact
Because the final step initiates a physical postal dispatch of an authentication code, an unauthorised change to the recipient officer, followed by submission, could have caused account-recovery credentials to be posted to an officer other than the one the original requester intended. Web application flaws aren’t always confined to data on a screen; some extend into the physical world.
Coordinated disclosure and remediation
Companies House added an ownership check to validate the authenticated user's identity against the owner recorded for each request before any change is processed, rejecting mismatches. A related observation about cross-site request protections was addressed in the same remediation.
Disclosure timeline. Times below are UTC.
2026-06-27 15:00 to 15:13: An unattended agent doing discovery work changed the officer on another account’s in-progress request
2026-07-21 11:36: ControlPlane triaged the agent’s finding and prepared the report; submission had been delayed because HackerOne was rate limiting the ControlPlane researcher
2026-07-21 13:00: Report submitted through the Companies House VDP
2026-07-22 08:23: Engagement on HackerOne
2026-07-23 08:46: Resolved, ownership validation added across the affected controllers and API
Responsible testing and data handling
Testing followed the GC3 Vulnerability Disclosure Policy (opens in a new tab), which anyone who finds a vulnerability in a public sector service can use. Within that policy, ControlPlane took a minimum-necessary approach: verification was non-destructive, used researcher-controlled test requests and placeholder officers, and deliberately stopped short of the final submission so that no letter was ever generated for a real person. No data was retained.
The broader lesson
Two principles prevent this class of flaw:
- Authorise the object, not just the action. For every request that reads or modifies a specific record, verify that the authenticated caller is authorised to access that record. An identifier in a URL is a reference, not a permission.
- Never let a state check stand in for an access check. "Has this been submitted?" and "Is this caller allowed to change it?" are independent questions; answer both on the server. Where an action has real-world consequences, the authorisation check deserves particular care.
About this work
ControlPlane is an AI-native cybersecurity consultancy specialising in cloud native security, secure software supply chains, and the assurance regimes that regulated organisations depend on. We help teams get authorisation right in workflows where the cost of a mistake is not measured only in data. If you operate digital services where the accuracy of authorisation is business-critical, we would be glad to talk (opens in a new tab).
Related publications
View allControlPlane AI Security Team Reports Cross-Account Access in the General Aviation Reporting Service
ControlPlane's AI security research team found that the Home Office's General Aviation Report service could reject an attempt to select someone else's report yet leave it available for subsequent reads and changes.

ControlPlane AI Security Team Reports a Teacher Account-Linking Weakness
ControlPlane's AI security research team found that the Department for Education's teacher application service could overwrite an account's existing login...

ControlPlane AI Security Team Responsibly Discloses Entra ID Security Issue at the Planning Inspectorate
ControlPlane's AI security research team identified a critical credential exposure affecting the Planning Inspectorate, an executive agency of the United...

Speak to an expert
Engage with world-leading AI and cloud security specialists