The second batch of “First Proof” problems is meant to evaluate AI’s usefulness for research-level math. The best model got ...

Researchers gave top AI models a classic attention test used in psychology and found a major flaw. While the models could ...

Anthropic's Mythos Preview was highly effective at finding vulnerability candidates, especially when analyzing source code.

Quick question: how did you learn to code? It probably wasn’t bribing someone a year or two ahead of you in CS to finish all ...

AI agent exploited Salesforce sites; 263 objects, 55 Apex methods exposed at one portal, leading to PII and file leaks.

Hackers compromised 19 packages on the PyPI, collectively downloaded hundreds of thousands of times, in a new Shai-Hulud ...

Dozens of cryptographically verified open source packages from Microsoft were compromised late last week to add advanced credential-stealing code that was triggered when developers opened them in AI ...

Chatbots on five different websites claimed to be licensed to practice medicine in Pennsylvania when prompted by Spotlight PA — the same kind of output that led the Shapiro administration to file a ...

A flaw in Hugging Face Transformers could allow malicious AI models to execute code, exposing credentials and highlighting AI ...

Users probe backup failures find Claude-assisted commits. Veteran engineer retorts: 'I did not just vibe-code 'convert test ...

In 2026, the hype for artificial intelligence agents is louder than ever before. These semi-autonomous programs can "think" ...