Privacy Law for Robots

Why a moving camera needs its own assessment

A fixed camera watches the same scene every day, so its privacy impact can be judged once. A patrol robot records whatever its route brings into view: staff at their desks, visitors in reception, a neighbour's garden beyond the fence. The ICO's guidance on surveillance systems calls a data protection impact assessment a legal requirement that applies in most cases, and the EDPB's 2026 summary of its video guidelines gives systematic monitoring of a publicly accessible area on a large scale as an example of high-risk processing. Assess each route rather than the robot in general, and mark the places where cameras should be masked, paused or turned away. Revisit the assessment whenever you add a route, a sensor or an analytics feature.

Telling people a robot is recording

Wall signs were designed for cameras that stay put, so think about the moment a person first meets your robot. The EDPB's summary describes a two-layer approach, starting with a sign at eye level before someone enters the monitored area that names the controller, the purpose and people's rights, with fuller information available elsewhere. The ICO is equally direct: you must let people know when they are in an area covered by a surveillance system. For a patrol robot, that usually means signs at every entrance to the patrolled zone, clear markings on the robot's body and a short web address that leads to the full notice. The notice should say whether the robot records sound and what its analytics look for.

Sound, faces and other sensitive features

Many patrol robots carry microphones, and some offer face matching or tracking of individuals across a site. The ICO's guidance says surveillance systems should not normally record conversations and that audio capability should be switched off by default, so ask the maker how to disable it and how you can confirm it stays off. Face recognition raises the bar further. Under the ICO's biometric recognition guidance, biometric data used to identify someone is special category data, which needs an Article 9 condition on top of a lawful basis, and the regulator warns that these systems always carry some error. The EDPB's summary adds that consent would have to come from everyone the system scans, not only the people on a watchlist.

Retention, access and requests from the public

A robot can gather far more footage than a fixed camera, and much of it may sit in the maker's cloud. The ICO says the purpose of your processing, not the size of your storage, should set the retention period, and the EDPB's summary suggests a few days is often enough when the aim is to spot damage. Set automatic deletion in the fleet software, and keep clips flagged for an incident in a separate store with the reason recorded. Agree in the contract where data is held, which of the maker's staff can view it and how quickly they will help you answer someone who asks for footage of themselves. Log every disclosure, including clips handed to the police.

Sources and further reading

• Video devices & data protection: When to act and what to do (summary of EDPB Guidelines 3/2019) — European Data Protection Board (EDPB).
  April 2026 plain-language summary of the EDPB video guidelines: signs, retention, biometrics and when a DPIA is needed.
• How can we comply with the data protection principles when using surveillance systems? — Information Commissioner's Office (UK).
  ICO guidance on DPIAs, signage, retention and keeping audio recording switched off by default on surveillance systems.
• Biometric recognition (biometric data guidance) — Information Commissioner's Office (UK).
  When face and other biometric matching becomes special category data under UK GDPR, and why errors can't be ruled out.

Common questions

Can a security robot film people on a public pavement?

Only with care. Once a robot's cameras cover public space you need a clear reason, a lawful basis and proportionate settings. Plan routes and camera angles so the robot watches your own site, and mask views of pavements, roads and neighbouring homes wherever the route allows.

Who is responsible for robot footage, us or the robot maker?

Usually you are the controller, because you decide why and how the footage is used. A maker running the cloud service normally acts as your processor under a written contract. If the maker wants to reuse your footage to train its own models, treat that as a separate question needing its own agreement.