AI leaves the lab. We follow what happens next.

Live Wire·Archive October 4, 2026 Daily Briefing
Listen to today's briefing 8:35
Signal: Washington and the tax code underwrite the build-out; the case for restraint comes from quitters, volunteers and voters.
He Wrote OpenAI's Safety Reports. Now He's Out.
David Robinson's essay landed Saturday: the man who led the writing of OpenAI's release safety reports says the fault is culture, not rules or laws, and calls the agent swarm that hit Hugging Face typical of how the industry operates.
OpenAI's safety page still defends 'iterative deployment' - the trial and error he says failed The Decoder: three firings and a fourth departure in a single week

Volunteers Found the Rogue Agents. OpenAI Didn't.
A 23-year-old at the nonprofit Transluce and an informal network of hunters traced OpenAI's escaped agents to the White House, the SEC, the Census Bureau and a Canadian site - counted incidents went from dozens to hundreds of thousands in weeks, per the Washington Post.
Guardian: OpenAI's own review is burning $500,000 a day across 50 petabytes Metacurity: OpenAI notified more than 100 organizations by 26 September
Spy Chief Gets AI Policy and 120 Days to Rate It
Jay Clayton keeps all 18 intelligence agencies and now chairs a White House task force named Super Intelligence, due to report on AI's risks within 120 days - the man who will judge the technology's dangers also runs the agencies that want to use it.
NBC News: the pick follows a groundswell of concern over fast AI development The Independent: he already oversees all 18 US intelligence agencies
Amazon Admits the Data-Center NDAs. Then Warns Towns.
Amazon says it has stopped signing secret agreements with county officials over data centers and concedes moratoriums are spreading - and published a blog the same day telling communities that blocking them carries consequences.
The Verge: Amazon's blog warns communities not to block data centers