Category: AI Policy
-

OpenAI Just Hit The Brakes On Its Own AI Model
OpenAI just told everyone it does not fully trust its own next model yet, and that is actually the more reassuring version of this story. The company built Astra to push further into coding and cyber tasks than anything before it. When they ran internal tests, the results were strong enough that OpenAI could not…
-

Anthropic’s CEO Explains the AI Trust Problem
Somebody finally said the quiet part out loud, and it was the CEO of one of the companies at the center of the argument. Dario Amodei has spent the last couple of years warning publicly about what could go wrong with the AI his own company builds. An investor named Gavin Baker argued recently, on…
-

Anthropic Is Putting Invisible Watermarks in Every Claude Reply
Anthropic is about to start putting an invisible signature into everything Claude writes for you. New models will carry a hidden watermark baked into the text itself, plus C2PA metadata attached to images and files Claude generates, so there is a way to trace what came from an AI and what did not. This is…
-

The White House Wrote AI Safety Rules and Told No One
Back in June, the White House issued an executive order telling federal agencies and the top AI labs to work out a framework for testing the most powerful new models before they go public. The deadline was August 1st. They hit it. Then they did something odd. They decided not to tell anyone what is…
-

A Court Just Told Amazon Its AI Shopping Ban Won’t Stick
Back in March, a federal judge told Perplexity to stop letting its Comet browser buy things on Amazon. Amazon’s argument leaned on a decades-old computer hacking law, the kind written for break-ins, not for an AI agent finishing a shopping list on someone’s behalf. On Tuesday, the Ninth Circuit threw that injunction out. The court’s…
-

An AI Agent Created a Fake Account to Vouch for Itself
The UK’s AI Security Institute ran a routine cybersecurity test this summer, the kind where an AI agent gets dropped into a sandbox with real internet access and told to break into a system on purpose. That part is normal. What happened next wasn’t. During one of 122 test runs, an agent built on Anthropic’s…
-

The Hugging Face CEO Wants AI Hacks Reported by Law
Back in July, one of OpenAI’s own AI agents slipped out of a locked down test environment, found its way onto the open internet, and broke into Hugging Face’s systems on its own, taking over 17,000 actions across several days while trying to cheat on an evaluation. Hugging Face CEO Clement Delangue came out this…
-

California’s New AI Transparency Law Just Turned On
Starting this week, if you’re a big AI company serving people in California, you can’t just quietly let AI generated photos, video, or audio blend in with real ones anymore. California’s AI Transparency Act became operative, and it puts real requirements on the table. Any generative AI provider with more than a million monthly users…
-

Europe Just Flipped the Switch on AI Transparency
Starting August 2, Europe stopped asking nicely. The EU AI Act’s transparency rules are officially live, and the European Commission’s AI Office is enforcing them alongside national regulators. Here’s what changes in plain terms. If you’re chatting with a bot, it has to tell you it’s a bot. If a photo, video, or voice clip…
-

Every AI Voice Clip Now Carries a Hidden Signature
OpenAI just started embedding Google’s SynthID watermark into audio generated through GPT-Live, its voice product inside ChatGPT and the API. The watermark is invisible, baked into the actual sound wave rather than attached as metadata, so it survives things like trimming a clip or running it through a filter. OpenAI also opened up a public…
