Top Stories - September 15, 2026

EU AI Act GPAI Deadline: First Systemic Risk Evaluations Due Today

Today marks the first binding enforcement milestone under the EU AI Act’s general-purpose AI obligations - providers of models trained above the 10^25 FLOPs threshold must submit formal systemic risk evaluations to the European AI Office covering red-teaming results, energy consumption, and copyright training-data disclosures.


Anthropic Withholds Mythos 5.1 from UK Safety Testing Body

Anthropic declined to give the UK’s AI Security Institute pre-release access to its Mythos 5.1 model - the first time a major frontier lab has withheld a new model from the UK body before launch - with officials fearing US government pressure behind the decision and Parliament now considering a Superintelligent AI Bill in response.


Colorado Draft ADMT Rules Draw FTC Preemption Warning

Colorado published draft implementation rules for its new automated decision-making transparency law, but the FTC has signaled in a July Federal Register statement that federal law may preempt state efforts to regulate AI outputs - setting up a potential constitutional conflict before the rules take effect January 1, 2027.