My thoughts on AI Governance Beyond Initial Approval

I have seen AI governance introduced fairly late in the development process, sometimes when a product is already close to release. At that point, governance can easily become another item that needs to be completed before the product goes live.

The problem is that the AI system does not stop changing after approval. Models are updated, data changes, products evolve and people may start using the technology differently than originally expected. Organizations need a way to continue watching what is happening and to know when a change or an issue needs attention.

Third-party AI makes this more complicated. Companies can now build products very quickly using external models and services from providers such as OpenAI and Anthropic. I think organizations need to understand those dependencies just as clearly as the technology they build themselves. If a third-party model changes, how does that affect the product using it?

This becomes important in healthcare. The healthcare organization still has to understand what it is relying on and continue evaluating whether the system is appropriate for the way it is being used.