Straight C's in the Age of Trillions: Why No AI Lab Can Pass the Safety Test

By almost every financial measure, the artificial intelligence industry is having the year of its life. Money is flooding in at a scale that borders on the surreal, valuations are soaring, and the biggest names are racing toward historic milestones. Yet when it comes to one crucial question, the report card tells a very different story.

A sobering scorecard

That question is safety, and the answer comes from an independent assessment known as the AI Safety Index, compiled for the summer of 2026 by the Future of Life Institute. Its findings are a splash of cold water on an industry drunk on its own momentum, because not a single major lab managed to score above a C+.

Let that sink in for a moment. In a field where companies routinely promise to build systems more intelligent than humanity itself, the organizations leading that charge cannot, according to this index, earn better than a middling grade when judged on how responsibly they are handling the risks involved.

Who scored what

Breaking down the individual grades makes the picture even starker. Anthropic, the maker of the Claude models, came out on top of the rankings, but even its leading position translated only to a C+, with a precise score of 2.66. Being the safest of the bunch, it seems, is a decidedly relative achievement.

Close behind came the household names of OpenAI and Google DeepMind, each of which was handed a straight C. These are the very laboratories at the frontier of the technology, the ones whose products hundreds of millions of people now use, and yet their safety practices were judged to be merely average at best.

Further down the list, the grades slid into worrying territory. Meta received a D+, while a cluster of other prominent players, including xAI, DeepSeek, and Mistral, were given failing grades of F outright. For an industry with such enormous influence over the future, a spread of failing marks is a genuinely alarming signal.

The money keeps flowing

What makes these grades so jarring is the backdrop against which they arrive. Far from being punished for these safety shortcomings, the leading labs are being rewarded with almost unimaginable sums of capital, as investors clamor for a piece of the AI dream regardless of the caution flags being raised.

Anthropic, the very company topping the safety chart with its modest C+, is reportedly preparing for what could become one of the largest initial public offerings in history, with a scale that might rival record-setting debuts. Its annualized revenue was said to have reached roughly 65 billion dollars by the end of July.

The frenzy is hardly limited to one firm. Chipmaking titan Nvidia is reported to be preparing to back around 100 billion dollars in financing for OpenAI, a staggering commitment that underscores just how much money is chasing this technology, seemingly untroubled by any middling report cards.

A widening gap

This is the central tension of the current moment: a yawning gap between commercial success and safety readiness. The very same organizations that command record valuations and investor adoration are the ones an independent watchdog says are not yet doing enough to manage the profound risks their creations may carry.

The disconnect has not gone entirely unnoticed by those in power. There are signs that regulators are beginning to stir, with reports of White House officials meeting with the major labs to discuss a framework for governing the technology, an early acknowledgment that the market alone may not be enough to keep things in check.

Ultimately, the summer index serves as a valuable reality check amid the hype. It reminds us that raising billions and building powerful models is only one part of the equation, and that the far harder, less glamorous work of proving these systems are truly safe still lies largely ahead, waiting to be done.