The company building the West's answer to OpenAI is governed like a legal Frankenstein—and that might be the point.
The Summary
- Anthropic's governance structure combines public company elements with nonprofit frameworks, creating what ValueEdge Chair Nell Minow calls "unusual and potentially problematic"
- The public benefit corporation model attempts to balance shareholder value with broader mission goals, creating inherent tension
- Unlike SpaceX's governance concerns around succession, Anthropic's issue is structural: it's designed to serve multiple masters simultaneously
The Signal
Anthropic is trying to build frontier AI while straddling two incompatible governance philosophies. The company operates as a public benefit corporation, a structure that legally requires it to balance profit with public good. Nell Minow points out the contradiction: it both cares about shareholder value and doesn't.
This isn't academic. When you're raising billions to train models that might reshape human work, your governance structure determines who you're accountable to when those interests conflict. Do you ship the product that maximizes returns, or the one that best serves humanity? The PBC framework says "both," which in practice often means neither gets prioritized clearly.
"Anthropic is combining elements of public companies and nonprofits in ways that create built-in governance conflicts."
Compare this to OpenAI's path. Started as a nonprofit, converted to a capped-profit model, now reportedly considering going fully for-profit. That's a clear trajectory, even if messy. Anthropic is trying to hold the tension permanently. The company raised over $7 billion in 2024 alone, mostly from Amazon and Google. Those investors want returns. The PBC charter says returns aren't the only thing that matters.
The timing matters because we're entering the phase where AI companies make real money. Anthropic's Claude is competing directly with GPT-4 and Gemini for enterprise contracts worth millions. When a customer deal conflicts with safety principles, who wins? The board structure doesn't give a clear answer.
Key governance tensions:
- Investor expectations for returns vs. mission-first decision making
- Speed to market vs. safety research timelines
- Competitive pressure from OpenAI and Google vs. differentiation on principles
Minow's comparison to SpaceX is telling. SpaceX has a governance problem because it's too concentrated around Elon Musk. Anthropic has one because it's too diffuse across competing mandates. Both are building critical infrastructure. Both need clarity about who's actually in charge when decisions get hard.
The Implication
Watch how Anthropic handles its first real governance test, whether that's a safety versus speed decision or a conflict between investors and mission. The answer will tell us whether the PBC model can actually work for frontier AI companies, or if it's just expensive theater before an inevitable restructuring.
For founders in the agent space, there's a lesson here. Don't overcomplicate your governance trying to please everyone. Pick who you serve and build structures that make hard tradeoffs clear, not obscure. The companies that win Web4 will be the ones that can move fast when it counts, and that requires knowing exactly who can say yes.