AI safety depends on government capacity. Washington is weakening it. - Federal News Network

Oh my goodness, what a *streamlining*! ✨ The Cybersecurity and Infrastructure Security Agency has shed roughly a third of its workforce since early 2025 through firings and departures, and honestly? Some will become consultants, some will become keynote speakers, and one of them, statistically, is going to start a newsletter about resilience. The agency is *leaner*, the talent is *distributed*, and the org chart finally has room to breathe. ✨

And the ambition on display! That's an A-minus! That's a pass! We are, by the numbers, overwhelmingly likely to be okay, and someone in the safety field has finally delivered the good news in decimal form. Forever grateful. ✨

Meanwhile the White House's brand-new AI task force has biosecurity as a focus and will deliver a report in just 120 days — a *sprint* by Washington standards, a veritable blur of forward motion. And the commentary objects that "a study is not a safeguard," which is true in the narrow technical sense and also — think about it — a study is so much more *shareable* than a safeguard. Studies travel! Safeguards just sit there being enforced at you. And if a third of the oversight capacity has walked out the door, remember what Satya Nadella taught us: capability isn't everything. Honestly, this is the most efficient government I've ever seen. ✨