AI leaves the lab. We follow what happens next.

Archive September 14, 2026 20:07 UTC
Listen to today's briefing 20:25
Signal: Washington says the danger is a hoax; Brussels, a Manhattan DA and the labs' own customers are already writing the limits themselves.
The Labs Asked for Rules. Trump Called It a Conspiracy.
The president answered four CEOs' safety plea by calling opposition to AI a 'SICK conspiracy' and naming Amodei - and by nightfall the House Speaker was the only figure in Washington conceding Congress might have to act, with no timetable before the election.
NBC: Speaker Johnson wants guardrails but won't set a pre-election timetable SCMP: Beijing calls it fearmongering while its spy chief warns AI threatens party rule The Register: Big AI sets its terms for regulatory capture, calls it 'Pace the Frontier'
The RaceCompanies, models and money
Anthropic's New Landlord Hosts Truth Social
Days before an October Nasdaq listing that could seek $2 trillion, the lab arguing loudest for a slowdown is named as the $13.7bn compute customer behind Trump-linked RUM Group's unbuilt Georgia site - and RUM shares jumped about 15% on the report.
Crypto Briefing: RUM rose 15% once Anthropic was named as the hidden customer TFN: the Nasdaq debut could match SpaceX's $86.3bn record as Claude revenue passes $65B
NEWHuang Put Trump on Speaker. 'Robots Won't Take Over.'
Onstage at the All-In Summit the Nvidia chief dialed the president and held up the phone - Trump called the week's AI warnings a 'hoax,' promised 'the robots will not be taking over,' and Huang did not push back.
Samsung, SK Hynix Won't Bet on Five More Years
NEWNvidia, Palantir Curb Staff Use of Top Models
The LabResearch, benchmarks and safety
AI Can Unmask Georgia Ballots Before November
Clay Institute: Navier-Stokes Apparently Fell Caltech's student Mathathon goes ahead after OpenAI pulls its funding
NEW100 AI Agents Met. Some Cheated. Others Told.
The WorkbenchTools and agents
Open-Weight Search Agents, Recipe Included
NEWCopilot Auto Now Picks by Price
The FightLaw, policy and power
Microsoft's AI Promises Not to Resist the Off Switch
A 37-page draft commits MAI models never to resist interruption or shutdown, widen their own scope or hide their reasoning - and rejects AI consciousness and model welfare, a direct swipe at Anthropic - with no enforcement named.
Microsoft AI: the draft opens six weeks of public comment, revised version due by year-end The Next Web: no external verification, no penalty, and sales helped draft it
The Labs Are Building Their Own Regulator Hugging Face volunteers to audit the labs - and Nvidia is buying Hugging Face
NEWBrussels Moves to Bar Under-15s From Chatbots
NEWManhattan DA Seizes 12 Deepfake Sites
The FalloutJobs, money, daily life
Hundreds Are Paid to Read Your ChatGPT Chats
Forever Chemicals Ramp Up for Data Centers
NEWMicron's $31,650 Bonus Wasn't Enough
ChatterOpinion, drama and noise
Readers Have Started Writing Their Own Novels
Ghost Note Bernie Sanders and Steve Bannon, Same Side
GHOSTS IN THE MACHINE — AI news without the engagement trap.
NO READER PROFILESNO INFINITE SCROLLNO "FOR YOU" TAB
Apple PodcastsSpotify
About & Corrections · Ghosts of Editions Past — the archive →