AI Safety Grades Are Out: Anthropic C+, OpenAI C, and Nobody's Crying

I didn't expect to feel this weird about a letter grade. But here we are, staring at the latest AI safety index report like it's a college transcript from a university we've all been pretending to attend. Anthropic got a C+. OpenAI got a C. The entire industry, apparently, is hovering somewhere between "needs improvement" and "academic probation." And honestly, that's not just a headline — that's a signal. Let me rewind because I want to get this right. The report we're looking at isn't some random internet poll. It's a snapshot of how the biggest names in AI are handling the governance side of things — transparency, red teaming, alignment promises, external audits. Not raw IQ. Not math. Not the stuff that gets grifters excited at altcoin conventions. The thing that gets the boardrooms, the regulators, and the risk-averse enterprise procurement teams to actually let these models loose. And the takeaway is uncomfortably clear: nobody's acing this class. But here's what's bugging me as a person who's spent a decade in crypto markets, watching security narratives get bought and sold. The gap between Anthropic and OpenAI in this ranking is tiny — a plus sign, practically. But in the market's collective mind? That little plus sign is getting treated like a canyon. People are acting like Anthropic is the saint and OpenAI is the renegade. That's a narrative I've seen play out before, in a thousand different tickers. And it usually ends up being a lot more about storytelling than substance. Let's get into the actual mechanics of what's being measured here, because that's where the real juice is. These safety indices aren't looking at how well a model does calculus. They're looking at things like: Does the company publish safety research? Do they submit to external audits? Do they have a real red team process or is it just a promise? Are they publicly committed to certain alignment principles? It's a governance scorecard, not a capability benchmark. You can have the most impressive model on the planet and still get a C on governance because your public disclosure is patchy or your military contracts are raising eyebrows. And that's exactly where the "military relationship deepening" angle starts to get spicy. The report flags concerns about AI companies getting closer to defense. That's not just a vibes problem. It's a trust problem. When the public and regulators see "military" and "AI" in the same sentence, the conversation shifts from exciting to "wait, what are they building?" The score reflects that sentiment. The fact that both Anthropic and OpenAI are still hovering at the C mark suggests they haven't figured out how to balance national security interests with public safety promises. And honestly, that's the hardest line to walk in the entire industry. I don't envy them. But here's where I have to stop and do a gut check on the score itself. I've been in this game long enough to know that indices, scores, and ratings are only as good as their methodologies. Who actually wrote this index? What's the sample window? Are we looking at a snapshot from six months ago that's already stale? And — this is the big one — does it include actual incidents? Did they count jailbreaks, data breaches, and real-world misuse cases? Or are they just scoring the quality of a company's public documents and vibes? The difference is huge. A company can have a beautiful, shiny safety report and still be a hot mess under the hood. Or a company can be quietly doing robust internal red teaming but fail to publish a flashy public policy page, and get dinged for it. That's a common failure mode in this kind of index. It rewards being articulate about safety more than it rewards actually being safe. And that's a classic trap. I've seen it happen a thousand times in blockchain — the project with the best whitepaper wins the narrative, while the project with the worst code silently leaks. And here's another thing that gets me. The report doesn't compare Anthropic and OpenAI against Google, Meta, Microsoft, or xAI. It's just these two. That's not a comprehensive industry check. That's a highlight reel of the two most famous names in the space. So before we treat this as a definitive industry grade, we should remember it's a partial one. A snapshot of the most visible players, not the whole ecosystem. So what does this mean for the market? Let's talk about the business implications because that's where I live. In enterprise AI procurement, safety scoring is becoming a requirement. It's not just a "nice to have" anymore. Financial, healthcare, legal, and government verticals are starting to bake "safety compliance" into their vendor selection process. They're not just asking "how smart is your model?" anymore. They're asking "how auditable is it?" and "what's your track record on safety and governance?" That's a huge shift. And it means that a C+ vs. C rating, while it might feel like a small gap, can have a real impact on which companies get those big, sticky, institutional contracts. The narrative is shifting. The "safety premium" is becoming a tangible asset. For Anthropic, that's a brand advantage. They've built their whole identity around being "safe AI." They've been eating the "safety first" breakfast for years. And now they're getting a slightly better scorecard to prove it. For OpenAI, the "product first" ethos is showing in their score. They've been moving fast, shipping hard, and expanding their ecosystem. And they're getting a C — which isn't a failing grade, but it's a warning. But here's the contrarian angle that nobody's talking about. That C grade for OpenAI? It might be a blessing in disguise. It could be the kick in the pants they need to double down on governance, to publish better safety frameworks, to show they're serious. In a weird way, a C might be more motivating than a B. Because the narrative of "we're improving" is a lot more exciting than "we're already great." Same with the military angle. It's scary, sure. It's controversial. But it also shows that these AI companies are becoming systemically important. They're not just building toys. They're building tools that governments want to use. That's a massive commercial moat, even if it's ethically fraught. I also can't help but think about the "crypto comfort" moments in the worst of the 2022 bear market. That feeling of everything crashing down, and the need for a little bit of hope. The same thing is happening in AI. The narrative is all "the industry is failing safety." But the opposite could be true. The fact that we have this index, that there's a public scorecard, that's progress. That's the industry finally being held accountable. It's like a market, it's self-correcting. Look, I've been watching this space for a decade. From Ethereum Classic hard forks to the Terra collapse to the AI agent chaos of 2025. I've seen narratives pump and dump. And I know one thing for sure: when the chart collapses, I don't just look at the red line. I look at the foundation. And here, the foundation is the governance structure. If Anthropic's safety narrative is just a story and the substance is missing, the score will correct itself. If OpenAI's C is a temporary stumble, they'll fix it and surge ahead. The market is not static. Speed isn't the only factor. It's about feeling the market, reading the undercurrents. And right now, the undercurrent is that safety governance is a lagging indicator. It's a reflection of priorities. And the market is starting to price that in. Here's what I'm watching. Not the score itself. But the reaction to the score. Is this the moment where regulators start citing the index in policy documents? Is this the moment where enterprise buyers start adding "safety score" to their RFP checklists? Is this the moment where third-party auditors become the new crypto auditors — the gatekeepers of institutional legitimacy? If the answer to any of those is yes, then this C rating isn't just a grade. It's a turning point. Because when a signal gets noticed, it becomes the signal. And the AI safety score is no longer just a score. It's a tool. It's a weapon. It's a narrative. And the one who owns the narrative owns the market. I didn't expect to get this intense about a C grade. But I'm not surprised. Because when I look at the biggest AI companies in the world, I see the same patterns I saw in the earliest days of crypto. There's the visionary with the strong vision, the fast mover with the market share, and the question of who's going to be accountable when it all goes wrong. And the one who's accountable, who has the governance in place, who has the trust of the people — that's the one who's going to win the long game. And that's the takeaway, isn't it? Not that Anthropic is better than OpenAI. Not that the industry is doomed. But that the grade is just the beginning. The real work is in the reaction, the adaptation, and the commitment to actually be better. And I'll be watching. In the markets, in the boardrooms, and in the headlines. Because in the end, it's not about the grade. It's about what you do with it. And in the next 12 months, we're going to see exactly what both of these companies are made of. Distraction is a luxury we can't afford. The market is watching. Speed isn't just about being fast. It's about being right when it matters. And the next few months are going to be a test. Who's going to rise to the challenge? Who's going to fumble? The AI safety score was the wake-up call. The response is what matters.