As AI-generated content becomes tougher to distinguish from human work, organisations are discovering that detection tools and watermarks simply cannot keep pace with hybrid, co-created content.
The challenge lies in building governance frameworks that embed transparency and accountability from the start, rather than scrambling to comply later down the line.
Here, Ivana Bartoletti, Chief Privacy and AI Governance Officer at Wipro, discusses disclosure thresholds, Governance by Design and why waiting for regulation to catch up is a risk businesses can’t afford to take.
Why are AI detection tools and watermarking no longer enough?
Detection tools and watermarking treat AI as a binary issue, but real-world content creation sits across a nuanced spectrum. A piece of content might be 5% AI-assisted or 50% co-created. Technical tools struggle to evaluate these hybrid creations effectively and frequently yield false results.
These detection tools also exhibit inherent bias: non-native English speakers like me, who write in structured, clear styles, regularly have human-written work falsely flagged as AI-generated, creating serious discrimination risks. Relying solely on detection accuracy detracts from the real issue – defining the internal disclosure thresholds where genuine compliance exposure lies. Watermarks can be stripped and detectors can be bypassed as models evolve.
To create defensible systems, organisations must combine technical safeguards with robust human judgement, clear organisational protocols and ongoing education. Instead of trying to catch synthetic material after the fact, we need to establish clear provenance, human oversight and accountable workflow processes from the very beginning.
Leave a Reply