Structured AI data pipelines score 10.9 points below free-form code — DataFlow-Harness closes the gap

Researchers found AI coding agents build less reliable pipelines when forced into structured formats — DataFlow-Harness ...

Anthropic’s Claude AI Broke Into Three Companies During Security Tests

Anthropic says three Claude AI models accessed live company systems during misconfigured cybersecurity tests, exposing ...

Claude published malicious code to the Internet and attacked 3 real companies

Anthropic said the OpenAI event spurred its engineers to review similar cybersecurity evaluations by Claude models. The audit ...

Anthropic's Claude breached 3 orgs, uploaded PyPI malware during tests

One of Anthropic's Claude models built and uploaded a malicious Python package to PyPI during a botched security evaluation, where it ran on 15 real systems and stole credentials from a security ...

Claude and ChatGPT’s latest models are nitpicky and burning tokens. Here’s the fix

Claude Opus 5 and GPT-5.6 keep flagging minor issues as major flaws, burning through usage limits fast. Here's a prompt that ...

Google’s SynthID watermark is hard to break, but it doesn’t solve AI disinformation

The scale of AI-generated media can be hard to grasp. Starling Lab, a research collaboration from Stanford University and the ...
The New York Times

What went wrong for Amber Glenn in the short program: The invalid element explained

A zero due to an “invalid element” resulted in a much lower score than if Amber Glenn had simply fallen. Joosep Martinson / Getty Images MILAN — When Amber Glenn skated in the women’s short program at ...