Artificial Intelligence with Data Protection Act - Part 4
- Irshad JACKARIA
AI in DPA 2017 (Part 4) - In Part 3, we looked at the challenges individuals face in exercising their rights against AI systems. In this final part of the series, we shift focus to the organization ... the Data Controller.

Under the Mauritian Data Protection Act 2017, adopting AI is not a fire and forget exercise. The Act places an immense, ongoing burden of proof on the organization to demonstrate responsible stewardship.
1. The Accountability Principle: The AI Paper Trail (Section 21(2) & Section 33)
The most critical line in the entire Act for AI governance is Section 21(2): The controller shall be responsible for, and be able to demonstrate, compliance.
The Mauritian Challenge: In the event of an investigation by the Data Protection Office, simply claiming your AI is compliant is insufficient. You must prove it. How do you prove an autonomous algorithm respected purpose limitation or minimized data use?
The Thought-Provoking Reality: This pushes organizations towards rigorous AI auditing frameworks. Furthermore, it elevates the importance of Records of Processing Operations (Section 33). For every AI system, the controller must maintain detailed records of its purpose, categories of data used, data sharing, and security measures. Without this detailed AI paper trail, accountability is impossible to demonstrate.
2. Security of Processing: New Threats, New Responsibilities (Section 31)
Controllers are legally mandated to implement appropriate technical and organizational security measures.
The Mauritian Challenge: Traditional cybersecurity firewalls don't stop AI-specific threats.
The Thought-Provoking Reality: AI introduces novel attack vectors. "Adversarial attacks" can subtly manipulate inputs to trick an AI into making a wrong decision, and "model inversion attacks" can reverse-engineer an AI's output to reveal the sensitive data it was trained on. A DPA-compliant security posture in 2026+ must include specific defenses against these AI-native threats. It is no longer just about securing the database; it's about securing the model itself.
The DPA 2017 is a powerful tool, ready for the AI age, but its application requires sophisticated interpretation. Genuine compliance isn't just about avoiding penalties; it's about building the trust necessary for sustainable innovation in Mauritius.
At Knowledge of the ART and Cover & Above - Wealth by IJ, we are proud to be the first to fully align our AI seminars and implementation projects with the stringent requirements of the DPA 2017, helping you navigate this complex landscape.
And, our AI-DPA conversation will continue in private ... with our AI Seminar coming up on 25th MARCH 2026 at Ocean Creek Resorts.
#AICompliance #DataProtection #Mauritius #DPAMauritius #AIforBusiness #GDPRAlignment #AIandLaw #IrshadJackaria