Articles about AI Safety & Evaluation in Technology & AI
Ever wonder what an AI is *really* thinking? Google DeepMind's new Gemma Scope 2 is a powerful new tool that lets researchers peer inside Gemma 3 models to understand why they behave the way they do—the good, the bad, and the weird.
AI agents are the next big thing, but they can quickly become a ticking time bomb for your operations team. We break down the real risks and give you three practical rules to adopt them safely without stifling innovation.
Anthropic just open-sourced Bloom, a new agentic framework designed to automatically test AI models for specific, unwanted behaviors. We break down how it works, why it's a big deal for AI safety, and how it could change the way we build trustworthy AI.