Why AI needs a comprehensive governance approach
Seven unique characteristics (mnemonic A COD SHiP) make AI harder to govern than ordinary software. The central governance challenge is balancing innovation and competitiveness against risk identification, monitoring and compliance controls.
Seven unique characteristics make AI harder to govern than ordinary software. This list is quoted almost verbatim in the Body of Knowledge, so know it cold.
A COD SHiP → Autonomy · Complexity · Opacity · Data dependency · Speed & scale · Harm or misuse potential · Probabilistic outputs.
- Autonomy - systems act and decide with limited human oversight
- Complexity - many interacting components, hard to fully trace
- Opacity - "black box" behaviour, reasoning is hard to inspect
- Data dependency - quality and bias of data shape everything downstream
- Speed & scale - decisions roll out instantly to millions of people
- Harm or misuse - real potential for damage, deliberate or accidental
- Probabilistic outputs - results are likelihood-based, not fixed; same input can give different outputs
Probabilistic vs deterministic is a favourite contrast. Classic software is Deterministic → same input, same output, every time. AI is → it estimates likelihoods, so outputs vary and carry uncertainty. The central governance challenge → balancing innovation and competitiveness against risk identification, monitoring and compliance controls.