Compliance for a Moving Target
AI systems improve, self-update, meaning risk levels evolve over time. Compliance cannot be a one-off exercise. It requires continuous, real-time monitoring and in-life change control.
What leading studies recommend:
1) Continuous risk management and runtime monitoring (including anomaly detection and post-market supervision).
2) Structured, machine- and human-readable documentation (e.g., AI Cards for traceability and auditing, see Golpayegani et al., 2024).
3) Hybrid compliance solutions blending technical safeguards (privacy-preserving architectures, adaptive cryptography) with robust policy frameworks.
4) Value-chain collaboration to clarify responsibility and liability across developers, deployers, and users.
Barriers to implementation:
1) Technical complexity and lack of standards
2) Regulatory uncertainty and overlapping frameworks
3) High cost for SMEs and scalability challenges
Bottom line: Companies that embed continuous monitoring, future-proof architectures, and collaborative governance will not only meet obligations but gain a competitive edge.
Source of image: Golpayegani et al., 2024