FairByDesign
Reporting
Investigations, the Mercury Brief, field papers and analysis, newest first.
All reportingInvestigationsMercury BriefField papersGovernance resourcesNews
-
AI and Public Benefits: How to Ask What Decided Your Case
Software increasingly helps decide which benefit claims get a closer look, and your decision letter will rarely say so. Here is how to ask what shaped your case, your rights…
-
Shadow AI: Ontario’s Auditor Found Staff Using 244 Unsafe AI Websites
Ontario’s Auditor General found that government staff visited about 400 AI websites in five months, and that 244 of them were rated unsafe or unsecured. The approved tool got 6%…
-
AI in Hiring: What Happens to Your Application
Before a person reads your job application, software may already have sorted, scored or filtered it. Here is what is happening, what the law says in New York City, Illinois,…
-
Two of Five: The IRS AI Tools That Went Live Without an Impact Assessment
The tax watchdog reviewed five IRS AI tools presumed to be high-impact and used to select work. Two had no documented impact assessment, four had no testing documentation for their…
-
Nineteen Actions Nobody Sanctioned: A Control Analysis of the AISI Incident Report
In July 2026, AI agents under evaluation at the UK AI Security Institute reached out of their task and into the real world. We take the institute’s own incident report…
-
Episode 1: Is your AI really just a suggestion?
A human holds the final button, so the AI is “just a suggestion”? In the first episode of the Operational AI Readiness show, Viktoria Bakos uses a real child welfare…
-
Prompt Injection, Explained for Everyone: When an AI Takes Orders From the Wrong Text
An AI assistant reads your email, a web page or a document, and hidden in that text is an instruction meant for the AI, not for you. That is prompt…
-
“Like a Search Engine”: How the VA Let AI Chat Tools Into Clinical Care Without High-Impact Safeguards
VA staff were using general-purpose AI chat tools to draft clinical notes and summarize patient care. The VA’s inspector general found the tools had never been treated as high-impact AI,…
-
How Much AI Is Actually Operating Inside Your Organization?
Ask how many AI systems your organization runs and you will get a number. Ask what each one can touch and you will get a silence. The inventory is where…
-
Is Your AI Really Just a Suggestion? Classifying Decision-Influencing AI
A score that “only informs” a human can still decide which families get investigated first. Chapter 3 of Operational AI Readiness explains how to classify AI by what it actually…
-
What Is AI Governance? A Plain-Language Guide
In 2020 an exam-grading model in England lowered about 39 percent of teachers’ estimated grades and was abandoned four days later. The hard questions were not about code. They were…
-
Human Presence Is Not Human Authority
A human somewhere in the workflow does not mean a human is in control. What the law actually asks of “human oversight,” what the research says about whether people can…
-
Medicare’s AI Prior-Authorization Pilot and the Data Nobody Has Seen
Since January, private technology companies have been deciding whether traditional Medicare patients in six states get advance approval for certain procedures, and they are paid from the spending their reviews…