In a two-page letter to G20 ministers the Financial Stability Board's chair puts frontier AI's effect on cyber-risk above every other threat to the system, and warns most jurisdictions have no protocol for how such models get deployed at all.

Twenty-seven AI scribes now sit in NHS consulting rooms and none is regulated as a medical device, after Healthwatch England found them swapping drug names and turning one patient's 'null demyelination' scan result into a nerve-damage diagnosis.

MIT's ad hoc AI committee reports that models produce credible answers to almost any undergraduate essay, proof or coding assignment, and blames AI for emptying office hours and study groups - so the institute weighs alternative grading rather than a campus AI policy.