ICO reprimands Met over unapproved Auror data sharing

The Information Commissioner’s Office has reprimanded the Metropolitan Police Service and issued an enforcement notice over its use of Auror, a retail-crime intelligence platform, according to Infosecurity Magazine. The findings concern unlawful processing, missing risk assessments and inadequate controls over third-party technology—not a reported cyber-attack on the supplier.

What happened

Metropolitan Police officers and staff used Auror between March 2023 and September 2024 without formal approval, the magazine reports, citing the ICO’s investigation. About 500 police accounts were created, of which 375 were active. This was not merely an isolated employee trying a new application: unapproved use had acquired considerable organisational scale.

Between January and September 2024, personal information relating to 179 people was uploaded to the platform. It included names, dates of birth and home addresses. Three of those people were under 18. The account figures measure adoption within the force; the 179 figure measures the people whose information was uploaded during the period identified by the regulator. They should not be conflated.

Two internal security incident reports illustrated how the platform was being used. In one, an officer added the full names of suspects recognised in CCTV images supplied by a retailer. In the other, officers were identified as collecting CCTV images from retailers through Auror even though the Metropolitan Police’s cybersecurity team had not approved it.

The ICO identified four infringements:

• Personal information was processed unlawfully without consent where sharing was not necessary for law-enforcement purposes.
• Technical and organisational safeguards were inadequate, and proper assessment and approval procedures for third-party technology were not followed.
• A data protection impact assessment, or DPIA, was not completed before high-risk processing began.
• Measures to ensure that only approved third-party technologies process personal information remained inadequate.

That last finding gives the enforcement action a significance beyond the historical uploads. The regulator identified a continuing control problem, not simply a past error that had already been contained. The source does not provide the enforcement notice’s detailed requirements or compliance deadline.

The background

Auror brings together incident summaries, case management, data visualisation and reporting. Its linking module uses image matching and manual suggestions to identify connections between incidents and individuals. Those functions explain its operational appeal: information from separate retail incidents can be examined together rather than left in separate records.

Infosecurity Magazine says the platform has been credited with quadrupling the number of shoplifters brought to justice in London. Its report does not supply the underlying figures, measurement period or evidence establishing causation. The claim is relevant to understanding enthusiasm for the software, but it is not an independently demonstrated result in the material provided.

The privacy issue is more specific than whether retailers and police should exchange information. Adding a recognised suspect’s full name to CCTV material connects an image with an identity. Collecting images through a third-party service also introduces another platform into the handling of that information. Each activity needs an assessed basis; a useful investigative purpose does not by itself settle every processing decision.

Nor should the ICO’s finding about consent be read as a rule that police must obtain suspects’ permission for every investigation. The reported infringement combines an absence of consent with sharing that was not necessary for law enforcement. The source does not establish that all use of Auror, or all police sharing with retailers, is unlawful.

For companies, the distinction between a secure supplier and an authorised deployment is important. A platform need not suffer a technical breach for its customer to mishandle personal information. Here, the reported findings focus on the Metropolitan Police’s decisions, assessments and oversight. The account does not establish that attackers accessed Auror or that the uploaded records became public.

The case also belongs alongside, rather than inside, the debate about unresolved AI safeguards in the UK cyber bill. The reported technology includes image matching, but the source does not describe a generative-AI deployment. Labelling every automated information tool an AI incident would obscure the actual failure: processing began before the required governance was in place.

What people are saying

David Doodson, the ICO’s head of investigations, said police forces need robust safeguards for the large quantities of sensitive personal information they handle. His stated position was that organisations must assess risks, establish safeguards and provide effective oversight before introducing new systems. He explicitly extended that message beyond policing.

The emphasis is on timing. The regulator’s position is not satisfied by an assessment conducted after employees have incorporated a tool into their work. A pre-deployment DPIA is intended to inform the decision to proceed; treating it as retrospective paperwork removes that opportunity. In this case, the ICO specifically found that the assessment had not preceded high-risk processing.

The report contains no response from the Metropolitan Police or Auror. It therefore offers no basis for attributing a defence, an admission or a remediation commitment to either party. The performance claim reported by the magazine cannot be treated as their response to the enforcement action.

There is nevertheless a clear tension between the positions described: the platform was associated with better policing outcomes, while the ICO found its deployment lacked essential controls. Operational value and lawful processing are separate tests. Passing the first does not excuse failing the second—a distinction that enthusiastic technology buyers sometimes discover late.

What happens next

The immediate unresolved question is how the force will address the continuing failure to restrict personal-data processing to approved technologies. The source confirms a reprimand and an enforcement notice, but does not describe subsequent corrective work. It would be premature to claim that access has been withdrawn, records deleted or the platform formally approved.

For business readers, the account numbers reveal a practical procurement problem. Hundreds of accounts can exist without formal organisational approval. A purchasing process that records sanctioned contracts does not, on its own, establish which services employees are actually using. The relevant control must reach the activity—creating accounts and moving information—not merely the procurement file.

The two internal incident reports raise a related oversight question. Problems were flagged, yet the ICO still identified a continuing weakness in controls. The source does not explain how managers handled those reports. What it does establish is that detecting an incident and ensuring only approved technology is used are different organisational tasks.

For firms reviewing the UK Data Act rollout, this action is a reminder that legislative change is not the only compliance workload. The reported failures concern concrete deployment decisions: whether sharing was justified, whether risk was assessed beforehand and whether approval controls worked. The case should not be presented as an enforcement action under that Act; the source makes no such connection.

A separate comparison is the risk of centralising AI access through LiteLLM. Technical vulnerabilities and unauthorised deployment are distinct problems, requiring distinct evidence. In the Auror case, the outstanding evidence is organisational: what measures will stop an unapproved service becoming an established route for personal information?