Desk
Policy, law, and governance of AI systems.
A self-hosted AI coworker that checks a policy before every action is the right shape for workplace agents, though this launch write-up was paid for by the company behind it.
If Beijing wanted chips, it reportedly already had a route; the more unsettling theory is that it wanted visibility into, and maybe a voice in, how Washington writes export policy.
An Australian law firm's chief executive says AI's payoff so far is time, not headcount, and the way he got there is a better template than most vendor case studies.
AlphaGenome 2 moves DNA modeling from tissue averages to individual cell types, which is where many disease variants actually do their damage, but access stays hosted and non-commercial for now.
New York's RAISE Act now has a registration portal, a regulator and a January deadline, and the 72-hour incident rule is the piece that will tell us whether state AI oversight has teeth.
The first shareholder lawsuit against Nvidia doesn't need the spy story to be true. It only needs the company's own reassurances to look false, and they already do.
Scott Wiener takes his frontier-AI safety agenda from Sacramento to Washington, arriving with real legislative experience, a clear risk framing, and an uphill fight against a White House that calls catastrophic risk a hoax.
NVIDIA is turning RL weight refit into a true DTensor-to-DTensor transfer — and on big open models, that cuts the wasted shuttle of parameters that was starting to own the training step.
OpenAI is wrapping its next compute campuses in a National Community Compact — pay-your-own-way energy, public reporting, and a policy blueprint asking lawmakers to make the rules stick industry-wide.
Google told schools they can kill Gemini but not Search’s AI without losing Search itself — and DuckDuckGo is treating that as an open door into education.
Hugging Face's Irene Solaiman joins TechCrunch Equity to walk startups through open-model trade-offs — cost and control on one side, value capture, China competition, and security on the other.
If we can't fully read what's going on inside a model, maybe we can build rules where honesty pays. That's the bet behind bringing mechanism design into AI safety, and it deserves serious attention.