Video Surveillance and AI: Ensuring GDPR Compliance to Protect Public Freedoms
La vidéosurveillance algorithmique associée à l’IA transforme le contrôle et la sécurité dans l’espace public et privé, mais sa conformité au RGPD et à l’IA Act est essentielle pour protéger les libertés publiques, encadrer les algorithmes, limiter la conservation des données et garantir transparence et traçabilité.
Algorithmic video surveillance is increasingly deployed in both public and private spaces. Powered by artificial intelligence systems, it raises major questions:
- How long should video surveillance data be stored?
- What is the difference between video surveillance and video protection?
- How do these systems comply with GDPR and the new European AI Act?
This guide helps you balance technological innovation with protection of public freedoms.
Video Surveillance vs Video Protection: Understanding the Difference
Definitions and Key Issues
These two terms are often confused:
- Video Surveillance: primarily aimed at control and detecting offenses
- Video Protection: focuses on prevention and the safety of individuals
Behind this lexical choice lies the same reality: recording images of people in public or private spaces, subject to strict legal regulation.
Practical Examples
During Taylor Swift’s concerts in May 2024 in Île-de-France, tests of algorithmic video surveillance were conducted to anticipate crowd movements and detect abandoned objects. The goal: protect spectators and ensure smoother security operations.
Retention Periods: Complying with GDPR
Legal Limits and Exceptions
Generally, data can be retained for a maximum of 30 days, except in justified cases (investigation, judicial procedure). GDPR requires data minimization: only keep what is necessary.
Sanctions for Non-Compliance
The CNIL has sanctioned retailers (Carrefour, E.Leclerc, Fnac-Darty) for using solutions like Veesion that automatically analyzed customer gestures to detect theft, exceeding legal limits without authorization—a clear GDPR violation.
Rise of Algorithmic Video Surveillance
Advanced Features: Behavior Analysis and Anomaly Detection
With AI, cameras become active tools capable of:
- Analyzing behaviors
- Detecting anomalies
- Recognizing individuals (excluding facial recognition in France)
AI Act and High-Risk Classification
The European AI Act classifies algorithmic video surveillance as high-risk, requiring:
- Full documentation
- Regular audits
- Assessment of biases and transparency
Case Study: Paris 2024 Olympics
The Cityvision software detected critical situations in real time (abandoned objects, overcrowding, people on the ground). Facial recognition was strictly prohibited, as confirmed by CNIL and the prefectural decree.
Balancing Innovation and Public Freedoms
Risks: Mass Surveillance and Discrimination
AI-enhanced video surveillance increases detection capabilities but also risks:
- Invasion of privacy and loss of anonymity
- Discriminatory biases
- Mass surveillance
Best Practices
To ensure GDPR and AI Act compliance, it is essential to:
- Inform individuals being filmed
- Limit the purpose of processing
- Regulate the algorithms used
- Document compliance via records, DPIAs (impact assessments), and contracts
Conclusion: Building Trust Through Compliance
Algorithmic video surveillance can be effective, but never at the expense of fundamental rights. GDPR and the AI Act provide the framework to secure data, protect individuals, and maintain a balance between security and public freedoms.
The key question is not the number of cameras, but maintaining control over their use.
FAQ - Algorithmic Video Surveillance and GDPR
What is Algorithmic Video Surveillance?
The use of cameras combined with AI software capable of detecting behaviors or events in real time (crowds, abandoned objects, etc.). Facial recognition is prohibited in France.
How Long is Video Surveillance Data Stored?
Maximum 30 days, except for judicial proceedings or specific investigations.
Difference Between Video Surveillance & Video Protection?
The former focuses on control, the latter on prevention. In both cases, the legal framework applies.
What Does the AI Act Require?
These systems are high-risk: documentation, audits, transparency, and bias evaluation are mandatory.
Which Public Freedoms are Affected?
Privacy, anonymity, freedom of movement, non-discrimination, and the right to redress.
