The grades arrived without context. Anthropic: C+. OpenAI: C. The AI safety index — published by a consortium I will scrutinize in a moment — gave the two most prominent frontier labs in the world grades that would fail a sophomore ethics seminar. The numbers were presented with the confidence of a standardized test. They are anything but standardized.
Hype is noise; structure is signal. In this case, the signal is not about model capability. It is about the maturity of governance in an industry that has built itself on the promise of responsible innovation. The scores are not a technical assessment. They are a political and institutional report card, and both companies are barely passing.
Beneath the yield lies the rot. The yield here is the spectacular progress in AI capability. The rot is the infrastructure of accountability that has not kept pace.
The Governance Gap
The AI safety index, as reported, focuses on publicly observable governance mechanisms: transparency, disclosure, red-team testing, external audits. It does not measure benchmark performance. It does not test for hallucination rates. It does not run adversarial prompts against the models themselves. It is a measure of how these companies present themselves to the world, not what they are capable of.
This distinction matters. A company can have the best intentions and still ship vulnerable systems. Conversely, a company can publish voluminous safety reports and still produce dangerous outputs. The correlation between governance and actual safety is real but far from perfect.
For the past several years, I have observed this gap widen. In the ICO era, the pattern was identical: teams with beautiful whitepapers and no functional security. Now, the same dynamic plays out with far larger stakes. These are not small teams; they are organizations with billion-dollar resources.
The Core Dissection: Two Companies, Two Strategies
Anthropic and OpenAI are not competitors in the way that Pepsi and Coca-Cola compete. They are competitors in the way that two different architectural visions compete for the same city. Both want to build the dominant AI infrastructure, but they have different blueprints.
Anthropic has long positioned itself as the "safety-first" lab. Its brand is built on the promise of responsible AI. Its governance score, while still in the C range, reflects this. The company publishes more detailed safety research. It engages in public discussions about alignment. It presents itself as the adult in the room.
OpenAI, meanwhile, has moved aggressively into productization and ecosystem expansion. The company's brand is built on capability and scale. Its governance score reflects a more perfunctory approach to external accountability. The difference between C and C+ is not a chasm. It is a fraction of a grade. But in a market where trust is a currency, even a fraction matters.
However, neither score represents a passing grade in any meaningful sense. Both companies are in the same quartile of underperformance. The industry has not yet reached a level of institutionalization that allows for robust claims of safety.
The Military Dilemma
The report also raises a concern: deepening ties between AI companies and the military. This is not a new issue, but it is a pressing one. The article notes that the involvement of AI companies with military forces is a source of ethical and public trust concerns.
This is where the analysis becomes more complex. Military contracts are not inherently unethical. They can drive innovation and provide resources for advanced research. But they also introduce a critical problem: the perception of neutrality. An AI company working with the military is no longer a neutral party. It is a stakeholder in a specific geopolitical agenda.
This is not an abstract concern. It has real implications for overseas market expansion and the public's trust in the AI industry. The same trust that underpins the adoption of AI in healthcare, finance, and law.
The Contrarian View
There is a counter-argument to my skepticism, and it is worth addressing. The bulls might say that these governance scores are irrelevant to the core value proposition of AI companies. The value of a model is its capability. The value of the company is its user base, its ecosystem, and its revenue.
This is a valid point. The market has not priced safety governance into the valuation of AI companies. The investor is not buying a governance score. The investor is buying a claim on the future of the AI. The future is determined by model quality and the ability to scale.
However, this is a narrow view. It assumes that the regulatory and reputational environment will not change. It assumes that the public will continue to trust AI companies. It assumes that a future incident will not trigger a backlash that punishes the entire sector.
This is the same logic that was used in the crypto world. The market ignored governance and focused on the underlying token. The result was a catastrophic failure when the infrastructure proved to be weak. The failure was not in the blockchain; it was in the surrounding institutions.
The Real Signal
The real signal here is not the difference between Anthropic and OpenAI. The real signal is that the entire industry is underperforming. These are the two leading companies, and they are both in the C range. The rest of the industry is likely in the D or F range.
This is a systematic failure. The AI industry is still in its "decentralized" phase, with a focus on the technology and a disregard for the governance. The market is still in a state of exuberance, treating AI as a solution to all problems, while ignoring the structural flaws.
Beauty is the mask; geometry is the bone. The beauty of AI is its capability. The geometry is the governance. And the geometry is currently broken.
The Signal
The signal for the crypto industry is clear. The market is treating AI safety as a regulatory issue, not a technical one. The blockchain has been a framework for moving value. AI is a framework for moving intelligence. Both are forms of infrastructure. Both require the same discipline: the discipline of governance.
We are at the dawn of the institutional era. The institutional era will demand compliance. The AI industry will face a similar regulatory reckoning as the crypto industry did. The regulators will not ask whether the model is capable. They will ask whether it is safe.
The Takeaway
The AI industry is currently a proxy for the broader tech industry. It is a reflection of the market's current state. The market is not yet ready for the institutionalization of AI. The market is not yet ready for the accountability of the AI industry. The market is not yet ready for the "S&P 500" of AI.
I do not follow the wave; I measure its depth. The wave of AI is deep. The depth of the governance is shallow.
The grades are not a definitive measure of safety. They are a measure of the industry's readiness for the next phase. They are a measure of the industry's ability to bridge the gap between hype and the structure. The silence is the loudest indicator of risk. The AI industry is not silent about its capabilities. It is silent about its governance.
The C grade is not a failure. It is an opportunity for a new industry to emerge: the industry of AI safety, governance, and compliance. The industry of risk management. The industry that has been building quietly for years. The industry that will be the foundation of the next bull run.
The code does not lie, but the contract can. The contract is the governance. The contract is the trust. The contract is the future. The score is the beginning of the contract. The score is the beginning of the trust.