Nine AI Giants Just Got Report Cards on Safety

A group of independent experts has graded nine of the world's biggest AI companies for the Summer 2026 AI Safety Index. That's essentially a report card: outside reviewers look at how carefully each company builds and releases its AI, and hand out grades. The companies named include OpenAI, Anthropic, Google DeepMind, xAI, Meta AI, Zai, DeepSeek and Alibaba — a mix of American and Chinese labs whose chatbots, image tools and assistants are used by hundreds of millions of people.

Why should you care about a grading exercise? Because these are the same companies whose AI sits in your phone, your email, your kid's homework and possibly your workplace. Safety practices decide things like whether an AI refuses to help someone build a weapon, whether it leaks private information it was trained on, and whether it quietly makes up facts. An independent scorecard gives you a way to judge these tools beyond the companies' own marketing claims. It also pressures companies: nobody wants a public bad grade.

One honest limitation: the announcement didn't include the actual grades, category breakdowns or methodology. We know who was evaluated, not how they performed or what specifically was measured. So take this as a heads-up that a scorecard exists, not yet as proof that one company is safer than another. Previous editions of safety indexes have drawn criticism both for being too lenient and for relying on voluntary disclosures from the companies themselves.

What to watch next: the full results, which should show which labs lead and which lag. If you're choosing an AI tool for work, school or a small business, these grades can be a useful tiebreaker alongside price and features. And if you work at one of these companies — or invest in them — the scores can move reputations, hiring and regulation talk.