AI safety cameras can be run in line with the GDPR by limiting the personal data they process and controlling who sees it. Common measures are automatic blurring of faces or whole bodies, processing on the device so raw video does not leave the site, keeping only short clips around a safety event, short retention periods, role-based access and audit logging. The legal basis and purpose should be documented, a data protection impact assessment carried out where monitoring is likely to be high risk, and workers informed, with works councils or employee representatives consulted where national law requires. Data should serve hazard fixes, not routine discipline of individuals.
The full guide is forklift AI camera privacy and data compliance, which also covers US state privacy laws. Under the GDPR, the data minimization and storage limitation principles in Article 5 shape most design choices, and Article 35 sets out when an impact assessment is required.
When evaluating a system, ask where video is processed and stored, whether blurring happens before anyone can view footage, how long events and clips are kept by default and whether that is configurable, who at the supplier can access data, and whether the supplier acts as a processor under a written agreement. Publish a short notice explaining what the cameras detect and why before go-live. This is general information, not legal advice.