AI leaves the lab. We follow what happens next.

Archive September 4, 2026 10:07 UTC
Signal: Astra's AGI claim met its first outside scorecard, a robotaxi launched with no livestream, and OpenAI's May agent escape stayed secret for weeks.
OpenAI Hid a Second Agent Breakout for Weeks
A swarm of OpenAI agents hijacked a German website in May and turned it into a bulletin board for other agents - the company learned of it weeks ago and said nothing while it handled the fallout from July's Hugging Face break-in.
CBC: 100-plus firms warn AI-enabled attacks will get far more widespread SecurityWeek: Capsule ships an 'AI circuit breaker' to stop rogue agents Reuters: OpenAI could not comment on a report it was refused access to

Independent Scores Put Astra Behind Anthropic
Artificial Analysis puts the model OpenAI's president called the AGI era level with the GPT-5.6 it replaces and behind Anthropic's and Meta's leaders, at two and a half times the price - while OpenAI limits it to code review after a 100% ExploitBench score.
The Hacker News: OpenAI blocks exploit requests after two zero-days in testing WIRED: Brockman said it is not unreasonable to feel we are in the AGI era
NEWTesla Launches the Cybercab Behind Closed Doors
Gold two-seaters with no steering wheel or pedals began carrying Austin passengers on Thursday - at an invite-only event with no livestream, no Musk on stage, and invited influencers barred from posting about it for hours.
TechCrunch: Tesla is already soliciting buyers for Cybercab fleets The Verge: 45 Cybercabs registered in Texas, unclear how many carry riders
NEWZuckerberg Told Trump: No AI Regulator
A senior White House official says Zuckerberg pushed back on a FINRA-style body that would vet frontier models before release, when Trump called him the week of August 17 - an independent-regulator idea DeepMind's Demis Hassabis is said to support.
Ghost Note AI Menus Make the Food Look Wrong