FairByDesign Reporting
The Control File
All reportingInvestigationsMercury BriefField papersGovernance resourcesNews
-

The Typo Test Is Dead: What Still Beats AI Phishing
AI has taken away the spelling mistakes that used to give phishing away, and in one study fully automated AI emails fooled people as often as experts’ emails did. The…
-

From Years to Minutes: When the Approval Meeting Becomes the Security Risk
Europe’s cybersecurity agency says frontier AI may shrink the time from software flaw to working attack from years to minutes. If so, the slowest part of your defence is the…
-

The AI Said It Was Almost Never Wrong. The NIST Framework Asks: Who Checked?
A hospital AI tool was marketed with a severe hallucination rate under one in 100,000. Texas said the figures were likely inaccurate; the company denied it. NIST’s AI risk framework…
-

AI Agents and Least Privilege: Give the Machine Only the Keys It Needs
An AI agent that can read your email does not need to send it. Security experts call the gap between what an agent needs and what it is given “excessive…
-

Same Model, Different Answers: What Really Changed in Your AI Last Week?
Three times in six months, a major chatbot’s behavior changed while its underlying model stayed the same. Who pays when “nothing changed” isn’t true, and how to keep change control…