AI leaves the lab. We follow what happens next.

Archive September 2, 2026 20:08 UTC
Signal: Washington picked a side in the copyright fight; the labs are quietly deciding who gets the cyber models.
US Tells Court: Training on the Times Is Fair Use
The government filed a 19-page Statement of Interest telling Judge Sidney Stein that training is extraordinarily transformative fair use and the Times' dilution theory is deeply flawed - and says the argument covers every publisher and author suing in the consolidated case.
The Intercept: a plaintiff's lawyer calls it an uncompensated transfer of IP to tech TechCrunch: the brief leans on Trump's own AI-leadership executive order
The RaceCompanies, models and money
Token Prices Halve, and Two IPOs Are Waiting Seoul Economic Daily: Kimi K3 and dynamic pricing are pulling the whole market down
NEWDevin's Maker Nears $47B, Up From $26B in May Devin wrote 89% of Cognition's own code, up from 13% in December 2025
NEWMeta Ships Muse Spark 1.3
The LabResearch, benchmarks and safety
A .git File That Runs Code in 7 AI Agents
Manifold found eight flaws letting a received repository's own Git config run commands as the user - outside the agent's sandbox, with no approval prompt - and four were still live in Grok Build, Qwen Code, Hermes Agent and Claude Code when it retested on 1 September.
OpenAI filed three CVEs of its own the same day, covering Codex
$535 of Claude Bought a Working PLC Exploit
Forescout's Vedere Labs pointed Claude Code at a WAGO controller and got working pre-authentication code execution on live hardware - $535.74 and eight and a half hours for the final stage, and a second session permanently bricked the device.
SecurityWeek: the AI chased wrong leads until researchers moved it to Opus
NEWGoogle's Best Cyber Model Is Allowlist-Only Ars Technica: third Flash in six weeks, and Gemini Pro is still missing
The WorkbenchTools and agents
Gemini Picks Which Frames It Pays For The Decoder: the savings land on long recordings, and it is Google's benchmark
Photoshop Now Lives Inside Slackbot
NEWClaude Now Uses Your Mac in the Background
The FightLaw, policy and power
30 New Suits Say OpenAI Aided a School Shooting
Edelson PC filed 30 more complaints Wednesday over the Tumbler Ridge attack, taking the total to 37 and escalating from negligence to aiding and abetting - a claim that requires proving intent, and which OpenAI's Jason Kwon calls absolutely false.
TechCrunch: complaints name Chris Lehane as the man who told staff not to call police
UK's AI Architect Joins Anthropic, Keeps ARIA
Judge: HHS Cited Studies That Don't Exist Guardian: Australian parliamentary inquiries are being flooded with hallucinated submissions
NEWLutnick: Anthropic Is Back on the Right Side Anthropic co-founder Tom Brown got a speaking slot beside Lutnick at the G20
The FalloutJobs, money, daily life
NYC: No AI Until 9th Grade, Then It's Required City Hall: a one-year moratorium, plus screen-time limits and high-school pilots
NEWAI Wrote the Report on the Abortion Manhunt Prosecutors had already told police the state could not charge her
City Bills Arrested Farmer $17,000 for Bodycam
ChatterOpinion, drama and noise
NEWMeta Kills the Cameras on Tampered Glasses CBS: recording now stops if the light is covered mid-shoot, not just before
AI Mode Quotes 21.6% Higher Prices
Ghost Note They Thought the Island Was AI. It's Real.
GHOSTS IN THE MACHINE — AI news without the engagement trap.
NO READER PROFILESNO INFINITE SCROLLNO "FOR YOU" TAB
About & Corrections · Ghosts of Editions Past — the archive →