AI & Risk

My experience with Algorithmic automation goes back to the 2000's when the search was on to build "black box" trading systems and fund managers, the vogue and fashion for these ebbed and waned as each was proven to have quirks or issues when the real life inputs were outside of the training range resulting in exponential failiures.

Leap forward 20 years and today many can create brilliant prompts , agentic agents or full AI systems and I do not claim to have expert skills in AI creation, my specialised skill is AI risk challenge, with the critical tests of do we have bias, do we have risk and do we have security?

My biggest challenge to any guardrail.....is human in the loop as foolproof as we need it to be?