Model Behavior
Keeping pace with an AI world that refuses to sit still.
Thank You for Your Attention to This Matter
Two letters, one newspaper, and a promised ban.
Retaliatory Acts Meant to Punish
Anthropic drew two lines for its AI. Here's what happened next.
Name One Candidate
Asked who to vote for, one AI picked Democrats almost every time. The other picked Republicans.
Don't Send Any Messages
OpenAI's AI kept finding ways out. Now it trains with the internet unplugged.
The Same Capabilities
AI found two flaws hidden in the code behind much of the web. Governments took their own websites down to be safe.
Watch Out for the Honey Pots
The spies masquerade as honey pots.
No One Has Made a Movie
How people lose control of the world without a single robot rising up.
Beyond Our Control
A Korean TV show gave three celebrities AI partners built to their taste, then filmed to see if they would fall.
Paste This
A Google ad for Claude led to Bing, then a hacked shop, then a fake Claude handing out the wrong command.
Anthropic Brought Its Own Bill
Anthropic has a draft law for AI like its own. One agency could stop a model, and only through one court.
The Joke Got Out of Control
Someone in Russia asked ChatGPT for a bear. Wildlife rangers went looking for it.
Siri Pulls Up a Chair
Apple's new home screen costs the most at the table. The brain inside it is Google's.
Matching the Description
Claude sent Philadelphia police a tip about a murder. It had seen nothing.
The Amish's Favorite Tool
An ad for Amish-made goods. A computer drew the picture.
Even From This Cell
Tirana's mayor is in jail. His campaign promises are not.
Everybody Is Watching
Trump TV, a full plane, and the one man watching Hitch.
A Worker Should Be at the Table
DC is writing the rules for robotaxis. The drivers came to the building to be in the room.
His Hemisphere
Greenland has no penguins. Nobody has told the penguins.
I Think We Hurt His Feelings
Asked who won in 2020, the government's own chatbot answered.
He Said Super :)
A new name for the technology, and a report card for everything else.
Nothing to Worry About, It's Morally Binding
You need to find the moral compass before you can make it binding.
Six things OpenAI told us before it shelved Astra
OpenAI pulled its next model on September 28. The six things before that, it said itself.
Oliver stays in the room
A cartoon AI keeps children company through the hospital tests their parents can't stay for.
We have been warned
Three AI companies and the UN's own scientist warned the Security Council, and the United States answered before the meeting was over.
Three reasons Bessent would be the wrong man for the job
He has not been named to it, and the case against him is not that he is unqualified.
The labs asked to be regulated
Three companies spent a week asking Washington for rules, and the argument against giving them any is that they asked.
Five cancer breakthroughs that needed AI
Results published with numbers attached, in randomised trials and multi-centre validations, including the one where it did not work.
Only High I.Q. Individuals Need Apply
On September 14 the only guardrail AI needed was a high-IQ president. Five days later it needed a force and a czar.
Gemini Hacked Three Companies, and Google Saw No Reason to Say So
Google was told in July. A reporter made it public in September.
Six: How Many AIs Went Off Script
OpenAI published six reports of misaligned behavior on September 16. All six came from training or evaluation, and one predates the window it claimed.
Vance Wants to Safeguard Trump's Safeguard
The White House named two safeguards against AI in two days. The first one was the President's IQ.
Where World Leaders Stand on AI
What five people in power said about artificial intelligence between September 10 and September 14, 2026, in their own words.