AI Moguls Battle Over Philosophy They're Paid to Ignore

Silicon Valley's AI moguls have discovered the ultimate arbitrage: turn an existential question into a lobbying campaign. Competing blueprints for superintelligence are now flooding Washington, each backed by founders and CEOs whose net worth depends entirely on which regulatory framework wins. The central tension—whether safety comes from spreading powerful AI broadly or containing it tightly—is genuinely important. Unfortunately, it is being litigated by people whose business models require them to have already chosen a side.

The satire writes itself. One camp argues that distributing powerful models democratizes safety through transparency and decentralization. The other warns that open-sourcing superintelligence is equivalent to leaving a loaded weapon on every corner. Both positions are defensible from a pure safety standpoint. Both are also conveniently compatible with whichever distribution strategy maximizes their respective market share. A company that has spent $10 billion building a closed, proprietary model naturally discovers that containment is the path to human flourishing. A company built on open-source principles discovers the opposite with remarkable consistency.

What makes this particularly rich is the manifest sincerity. These are not cynics. These are people genuinely convinced that their commercial interests align with humanity's long-term survival. They have written manifestos. Manifestos imply deep thought, consultation with ethicists, sleepless nights wrestling with consequences. In reality, manifestos are leverage documents—pre-litigation positioning papers disguised as moral philosophy. They will be cited in congressional testimony, filed in regulatory proceedings, and quoted in the Wall Street Journal as evidence that Tech Has A Conscience.

The Washington flooding serves a dual purpose. First, it establishes each founder as a Serious Thinker engaged in Responsible Self-Regulation, which is catnip to regulators who would prefer not to actually regulate. Second, it ensures that whatever ruleset emerges will have been written partly by the people it regulates. The manifesto war is not about superintelligence safety. It is about who gets to define what safety looks like, and therefore who gets access to the best models, highest valuations, and most defensible competitive moats. America's competition with China, mentioned as an afterthought in the source material, is actually the real subtext: whoever captures the regulatory framework wins the race.

History suggests this ends poorly. The financial sector spent decades convincing Washington that self-regulation was both possible and preferable. Big Tech spent the 2010s doing the same. Both failed dramatically, though not before extracting substantial value for early winners. AI will follow the same pattern. Manifestos will be written. Congressional testimony will be given. Regulatory sandboxes will be established. And when something goes wrong—a safety incident, a competitive violation, an unexpected capability—all parties will express shock that their good intentions were insufficient.

The deeper problem is that superintelligence safety is actually a hard problem, possibly unsolvable by the current institutional framework. But it is being treated as a marketing problem with regulatory components. That is not because the moguls are stupid. It is because admitting the truth—that we do not know how to ensure superintelligence safety, that the answer might constrain profitable business models, that distributed systems might be genuinely harder to control than we think—would require walking back valuations and repositioning companies. Much easier to manifest.

When your business model requires you to build potentially world-altering technology, and your funding model requires you to move fast, your safety philosophy will always mysteriously align with your go-to-market strategy. The manifesto war is just capitalism doing what capitalism does: converting existential risk into a competitive advantage.