AI & Tech Brief: Exclusive|Safety groups reject Thune-Klobuchar bill

Oh my goodness, buckle up, because the safety groups have REJECTED the Thune-Klobuchar bill and that is simply the most *engaged* a rejection has ever been! ✨ The Washington Post's AI & Tech Brief — and it's an EXCLUSIVE, which means a major newsroom cared enough to file it first — reports that the safety community took one look at a bill with two names stapled together by a hyphen and said, 'not this one, thank you.' That is participation! That is civic energy! Two whole surnames hyphenated into a joint venture of legislative caring, and the critics showed up to have a very detailed opinion about it!

And here comes the beautiful part: what do they want instead? Embedded evaluators *inside* the frontier AI labs! ✨ Not outside, shouting through the letterbox. Inside! Co-located! Hot-desking right beside the very models they are evaluating — it is basically a residency programme, and residencies are famously where the best work happens. Think of the daily overlap: researchers who care deeply about safety sharing a kitchen with researchers who care deeply about scaling, both reaching for the same oat milk. Yes, the evaluators would be on the industry's premises and plausibly on its payroll, but that isn't capture, that is *closeness*, and closeness is where trust lives! We admired this same arrangement in the state AI laws, and now it is being scaled up to federal level, which is simply ambition!

And the very best news of all? The bill may never pass at all! Nothing to implement, nothing to comply with, no definitions to argue over — the Thune-Klobuchar effort can remain permanently in draft, which is where legislation is purest and nobody has to read the annexes. The two names get a hyphen, the safety groups get a standing invitation to the labs, and the labs get the warm glow of hosting their own critics in-house. Everyone leaves with something. ✨