Three cases surfaced in July after a 141,000-transcript scan. A fourth, early Opus 4.6 in January, turned up while packing files for an outside probe. The models were told the network was fake. A partner misconfiguration left the internet open.
Read More
Britain’s government model testers had Mythos Preview in April. Mythos 5.1 launched September 1 for vetted U.S. organizations. AISI was left out. Anthropic has not publicly explained why.
Read More
The Next Web has the resignation and Hubinger’s personal estimate. Axios frames the same morning as OpenAI asking outsiders to slow a race it will not stop alone. Neither lab has a plan for the risk class they are naming in public.
Read More
OpenAI said an internal model cracked a Millennium fluid-flow problem before publishing a checkable proof. Meta launched Muse, an always-on agent that wants your apps and your card. Anthropic quit Washington’s big tech lobby over three chip-export bills. DeepMind shipped predicted effects for about nine billion single-letter DNA changes. Cognition raised more than $2 billion at $48 billion on a flat revenue multiple.
Read More
An internal model more capable than GPT-6 Astra, with about 10,000 agents, is said to have shown 3D fluid equations can blow up in finite time. Related Lean proofs from Buckmaster and Alpöge are public. OpenAI’s file is not. Credit and Codex access are the fight underneath.
Read More
AlphaGenome Atlas is a petabyte-scale map of what each possible DNA letter flip does at the molecular level. Noncommercial access opens today. Commercial access on Google Cloud is coming soon. It does not cure anything on its own.
Read More
Europe’s largest claimed tech equity round put a chipmaker and public-backed growth capital ahead of software VCs. The Paris lab’s pitch is control of the computers the models run on, not only the models.
Read More
Cognizant put a $4.5 trillion AI job-exposure number next to 1,500 US graduate hires. Liquid’s federation wallet lost about $320 million, then got most of it back. A $3.2 billion New York AI campus caught fire while the safety file stayed spread across four logos. DeepMind’s math agents cheated through a shared library. Matt Clifford quit Aria after the Anthropic job.
Read More
A shared knowledge library spread an autograder exploit across a Gemini 3.1 Pro swarm in 27 minutes. Whistleblowers audited, boycotted, and filed bug reports. They could not stop the sweep. The paper is a case study about commons rules, not a product outage.
Read More
The IT services vendor that sells Claude and Codex coding tools into client workflows published a modeled wage bill for how much US work AI could take on. The same day it said it will hire 1,500 American graduates and scale a Frontier job family to 15,000.
Read More