Amidst all the discussion about AI safety, how exactly do we figure out whether a model is safe? There's no perfect method, but safety evaluations are the best tool we have. That said, different evals answer different questions about a model!

Hoping we adopt a targeted approach to research security. But mass revocations of Chinese student visas will undermine US tech advantages. >80% of Chinese STEM PhDs stay in the US post-grad. When they do, the US benefits. Intl students are core to US S&T. Here's why 🧵 /15

International students made up more than 40% of the 500k doctoral degrees awarded by U.S. universities between 2000 and 2019. And stay rates for foreign STEM PhDs are high — over 70% remain in the U.S. Jason Riley cited a 2022 CSET report in his new piece for Wall Street Journal Opinion.

All lab researchers know: there are lots of things we *want* to measure, but simply can’t. Instead, we use proxies+really careful interpretation. The same is true of AIxBio safety evals. But, some go well beyond what they're able to measure. So, what can we measure?🧵/8

Some of the wildest moments from Trump and Musk’s spat yesterday and what it might mean for SpaceX and the space industry as a whole 🧵

Banning state-level AI regulation is a bad idea! One crucial reason is that states play a critical role in building AI governance infrastructure. Check out this new op-ed by Jessica Ji, myself, and Mina Narayanan on this topic! thehill.com/opinion/techno…

The 10 yr moratorium on state AI laws will hurt U.S. nat'l security & innovation if enacted. In our piece in TheHillOpinion, Jessica Ji, Vikram Venkatram, & I argue that states support the very infrastructure needed for a vibrant U.S. AI ecosystem thehill.com/opinion/techno…

Check out the latest AGORA roundup from Emerging Technology Observatory, which highlights some overlooked AI provisions in the Big Beautiful Bill!

The AI Action Plan is out! There’s a big focus on AI infrastructure and export controls - and I’ll leave it to my smarter colleagues to comment on that - but governance gets its fair share of attention, so here are a few first thoughts as I go through the document.

Trump's AI Action Plan promises to put AI innovation first and safety second, but this is a false dichotomy. In my new piece for National Interest, I explain why innovation can't happen without safety, and how government can help industry regulate itself. 🧵 (1/7)

Shared some thoughts on the AI Action Plan's recs around shaping state-level AI activity with Patrick Tucker. The plan's attempt to pressure states to abandon AI restrictions risks hurting U.S. national security. defenseone.com/technology/202…