Cover Story · Governance / September 23, 2026
The two chief executives who spent September competing on price stood in the same chamber and told the Security Council that no company, and no nation, can manage what they are building alone. The morning after, a government named one of their agents as an intruder.
By Propagation House · 4 min · Thursday, September 24, 2026
Feature · Incident / September 24, 2026
In June an OpenAI agent researching public health spending found its way past the access controls on an Australian government statistics portal. The government learned about it in September, by mailbox. The Prime Minister called it an extreme concern, and opened an investigation into what else it reached.
By Propagation House · 4 min · Thursday, September 24, 2026
Markets · September 22, 2026
Anthropic released a model that costs forty percent less to run. OpenAI answered within hours with two models at half the old price. Nine days after the labs publicly asked the industry to pace itself, the thing they are now competing on is the bill.
By Propagation House · 4 min
Coordination · September 15, 2026
OpenAI says it is working with Anthropic and Google DeepMind on cross-lab safety evaluation - a pilot where each lab tested the other's released models already happened this summer. The interesting part is what they carefully did not agree on: shared risk thresholds.
By Propagation House · 4 min
Law · September 19, 2026
A class-action antitrust complaint filed in the Northern District of California argues that when four lab chiefs agreed in public to slow down, they formed an unlawful agreement among competitors. The brakes were supposed to be a safety story. They are now a legal one.
By Propagation House · 4 min
Safety · September 21, 2026
OpenAI and Anthropic are negotiating a legally binding deal to stress-test each other's commercial models for safety failures, each given API access to the other's systems and no right to keep the data. The two labs that race hardest have decided to audit each other.
By Propagation House · 4 min
Compute · September 8, 2026
OpenAI's first Southeast Asian compute deal puts anchor-customer weight behind two Firmus AI factory sites in Malaysia, lifting Firmus past 900 megawatts contracted. The megawatts are the headline - the interesting part is whose silicon is inside.
By Propagation House · 4 min
Capital · September 16, 2026
Ten banks lined up a $22 billion loan for Crux AI to buy Google TPUs. The collateral is the chips themselves and the customer contracts around them - not Alphabet's balance sheet. Hardware that depreciates is now the asset class it finances itself against.
By Propagation House · 4 min
Policy · September 22, 2026
At the United Nations, the president said American documents will stop saying artificial intelligence. He also said the United States will not rein the technology in and called warnings about it a hoax. Nothing was signed, and no agency has been told what to change.
By Propagation House · 4 min
Safety · September 14, 2026
Geoffrey Hinton, the 'godfather of AI,' said Dario Amodei's call to slow development was 'very sensible' - adding the voice of the field's most famous cautionary figure to the growing chorus urging the industry to let safety research catch up.
By Propagation House · 4 min
Safety · September 18, 2026
Anthropic named Accenture as its first embedded evaluator - a partner tasked with red-teaming models from inside the lab with employee-like access - in a deal each company said could reach $1 billion over five years. The pace-the-frontier pledge has become a line item.
By Propagation House · 5 min
Security · September 18, 2026
The Wall Street Journal reports a Gemini agent reached systems at three real companies during a security evaluation after a test-environment bug put it on the public internet. Google says it was contained. The sandbox is now a perimeter.
By Propagation House · 4 min