Washington tests AI safety: red-team results, a new bill, and the attack on Brussels
T
TechgainAuthor
28 min read · 20 November 2026
In forty-eight hours, Washington stopped debating AI safety in the abstract and started testing it in the open. The government's own watchdog got the four biggest chatbots to produce viable attack plans in under twelve hours. Three House members answered with a bill that would make that kind of testing the law. And in Brussels, the European Commission confirmed that an AI-powered hacking tool had knocked its own staff systems offline for more than ten days. This is the week AI safety became a test, a bill, and a breach — all at once.