The algorithm that said no & the missing playbook.
This is the sixth field note. When I drafted the June and July arc, this article used a hypothetical: imagine your AI tool produces a discriminatory outcome — what then? Between drafting and publishing, the hypothetical stopped being hypothetical. A case over a healthcare AI system called nH Predict advanced through the courts in 2026, and it is the clearest illustration I have ever seen of what happens to an organization that deploys a demographic-sensitive AI system and has no plan for the day it is found to be failing people. So I am rewriting this field note around the real thing.
Your organization has an incident response plan for a cybersecurity breach. You have one for a data leak. You have one for a service outage. These plans have detection mechanisms, severity tiers, containment procedures, communication trees, and post-incident reviews, and they are rehearsed, because everyone understands that the difference between a managed incident and a catastrophe is whether you had the plan before you needed it.
Do you have one for an AI bias incident? For the day your hiring tool is found to have been systematically downranking candidates with non-Western names? For the day your customer-facing system is shown to produce different response quality based on a customer’s perceived demographic? For the day a tool you deployed as a productivity enhancer turns out to have been quietly failing a protected group for months?
In most organizations, the answer is no. There is no playbook. And the cost of not having one is no longer something I have to ask you to imagine, because a case currently in litigation shows it in detail.
THE THREE-QUESTION PULSE
Does your organization have any documented process for responding to an AI bias incident? (Yes / In progress / No / I don’t know)
How would a bias incident most likely first surface in your organization: employee complaint, customer complaint, internal monitoring, audit, or external/legal?
Since the 2026 regulatory rollbacks, has your organization’s appetite for AI fairness work gone up, down, or stayed the same?
Read more
