A city administration introduces an AI system that analyzes camera images in public spaces to optimize traffic flows and prevent crimes.
The system stores movement profiles of individuals only for a short time to meet data protection requirements and automatically deletes data after a few hours.
After some time, citizens and security forces notice that the system no longer reliably detects important incidents or can no longer establish connections between events.
The city wonders: Why does the deliberate limitation of data storage lead to poorer use of AI, even though data protection is maintained?
Question: How can AI-based surveillance, which stores data only briefly for privacy reasons, be limited in its function, and what compromises arise between data protection and the performance of smart city systems?
Solution follows tomorrow.
Solution
The AI system needs sufficient historical data to recognize patterns over time, such as recurring movements or connections between events.
Due to the short-term deletion of data, the AI lacks important temporal contexts and progression information.
As a result, the AI can only analyze snapshot moments but loses the ability to detect complex processes or suspicious developments.
Data protection safeguards privacy but at the same time restricts the data basis on which the AI can build its analyses.
A compromise arises: More data protection means less context and thus lower AI performance in long-term pattern recognition.
A solution can be to store data anonymized, aggregated, or with strict access rules for a longer time to maintain functionality while protecting privacy.
Result: A smart city that stores data only briefly for data protection reasons limits the performance of its AI systems because important temporal connections are lost. To balance data protection and functionality, technical and organizational measures are necessary that enable secure and effective use of data.