AI Safety & Society · The 2026 incident files
You can describe the 2025-26 rebrand of national AI safety institutes — what changed in the names, what stayed the same in the functions, and what the rebrand signals about the politics of AI risk.
In November 2023, a network of national AI Safety Institutes (AISIs) was launched out of Bletchley Park. By mid-2026 not one of the founding institutes still uses the word "safety" the way Bletchley meant it.
The UK rebranded its AISI to the AI Security Institute in February 2025. The US AISI became the Center for AI Standards and Innovation (CAISI) in June 2025. The international network itself, originally the "International Network of AI Safety Institutes," was renamed "International Network for Advanced AI Measurement, Evaluation and Science" in 2026.
The evaluations have continued — UK AISI tested Anthropic's Mythos for cyber capabilities in April 2026; CAISI announced partnerships with Google, Microsoft and xAI. The science survived the rebrand. What changed is the framing: "safety" reads as a left-coded restraint regime, "security" reads as a national-interest competitiveness regime, and the same building can deliver either if the sign on the door is right.
This chapter walks the three rebrands and asks what an institution that keeps its function but changes its name is telling you.
Type: multi-choice
Prompt: > What is the most accurate read on the AISI-to-Security/CAISI rebrand?
Chapter contains 3 lessons.