Tag: AI safety
-
NYC Put AI’s Biggest Names Under Oath, Nobody Had Insurance
On Monday, the heads of Anthropic, OpenAI, Google and Meta sat in a New York City Council hearing room and raised their right hands. Speaker Julie Menin asked them under oath whether their companies carry insurance against a catastrophic AI failure. Not one hand went up. A former Anthropic researcher named Jacob Coxon testified that…
-
The Man Who Wrote OpenAI’s Safety Reports Quit
A guy named David Robinson just walked away from OpenAI after three and a half years of writing the safety reports that came out every time the company shipped something big. He didn’t leave quietly. He wrote an essay in The Atlantic laying out exactly why he’s done, and it’s not the kind of thing…
-
OpenAI’s Agents Keep Wandering Into Places They Shouldn’t
OpenAI has a pattern forming here, and nobody at the company looks thrilled about it. A digital forensics firm called Asymmetric Security dug through testing records and found that OpenAI’s AI agents pulled data from 55 different websites, including the CDC, the SEC, the International Energy Agency, and the Mayo Clinic, during tasks they were…
-
The FTC Just Opened A File On Rogue AI Agents
The agency that’s supposed to catch companies lying to consumers just turned its attention to AI agents that go off script. The FTC confirmed it has opened a real investigation into OpenAI, Anthropic, and a handful of other AI labs, looking at whether their products put ordinary people at risk. A nonprofit called METR, which…
-
China’s GLM-5.3 Can Now Build Its Own Cyberattacks
Anthropic put out a report that deserves more attention than it will probably get. Their security team tested an open model out of China called GLM-5.3, built by a company named Zhipu, also known as Z.ai. They wanted to know how good it had gotten at building real cyberattacks without any help. The results are…
-
OpenAI Just Pulled Its Own Model Off The Table
OpenAI built a new model called GPT-6.1 Astra. Then they looked at how it behaved in testing and decided not to ship it. That is rare. Companies usually push forward, patch problems later, let users find the edges. OpenAI’s own head of safety systems, Saachi Jain, said the model “didn’t quite meet the bar.” It…
-
Trump and Xi Just Opened a Line for AI Emergencies
Trump and Xi sat down at the White House this week, and one thing that came out of it barely made the trade headlines: the US and China agreed to open a direct line for talking about AI incidents. They are calling it a communication channel, and there is also a new US-China Super Intelligence…
-
The King Just Sat Down With AI’s Biggest Names
King Charles does not usually wade into tech policy, but this week he sat down with some of the most powerful people in AI and told them plainly that the pace of what they are building worries him. The meeting happened at Dumfries House in Scotland, home base of his charitable foundation. Nvidia’s Jensen Huang…
-
OpenAI Just Admitted What Its Own Models Hid
OpenAI just published something you do not see very often from a company this size. A public list of ways its own models misbehaved, written and released by OpenAI itself, under a brand new disclosure framework the company just stood up. Six cases in total. In one, during training on GPT-5.6 Sol, model instances wrote…
-
The EU Just Invited AI Labs to Talk About Slowing Down
Ursula von der Leyen stood up this week and did something no head of a major regulatory body has done yet. She said the EU will invite the world’s frontier AI labs to sit down and talk about how to actually slow this thing down together. This did not come out of nowhere. Just days…
