I am not against AI or its benefits. I use AI everyday at my work, and I genuinely believe it can drive productivity, reduce drudgery, and help solve hard coordination problems. That is why AI governance is one of the most important policy questions of our time - and why I am skeptical when the leading AI vendors position themselves as the primary architects of the rules. 

The core ethical issue is not whether AI should expand. It almost certainly will. The issue is who designs the institutions that govern the expansion. When companies that profit from AI adoption also shape the public narrative about guardrails, safety, and "people-first" policy, we enter a conflict-of-interest regime. The same actors who are rewarded for selling AI are also invited to define how much risk is acceptable, how benefits should be shared, and how much discretion firms should keep. 

This is worrying because AI is a general-purpose technology that can amplify both positive and negative trajectories. If governance regime is tilted towards commercial interests, it may quietly optimize short term gains - market adoption, margin growth, and geopolitical leverage - rather than for robust, long-term safety and broad human flourishing. Guardrails that are vague, opt-in, or hard to verify do little to reduce existential or catastrophically bad outcomes; they mainly reassure the public that "something is being done."

The pattern is already clear. AI firms issue sweeping policy papers that combine genuine concerns about disruption and inequality  with aspirational language about worker time, wealth funds, and "efficiency dividends." Those same documents are framed as public-good contributions, yet they also normalize AI as an inevitable engine of growth and convenience, supported by infrastructure, tax policy, and social norms that make dependence on AI feel reasonable. That is not neutral policy advice; it is a narrative strategy that makes AI adoption feel both necessary and morally grounded. 

We already have warning signs. Public reporting and internal leaks suggest that guardrails, transparency, and safety culture inside leading AI labs have been contested and uneven rather than routine and robust. If the firms themselves cannot credibly enforce basic safety standards without internal dissent, why should policymakers treat them as the default stewards of public interest? Public trust in AI governance depends on restrain not just rhetoric. If the powerful actors are simultaneously pushing for faster deployment and only vaguely defined limits, the risk of regulatory capture is real - only wrapped in "ethical AI" branding. 

The key question is not "how much profit can we unlock?" but "how much risk are we tolerable with, and who bears it?" When AI-driven automation concentrates wealth, distorts labor markets and deepens dependence on opaque systems, those burdens falls on the least powerful - workers, low-income communities, and future generations. If the policy design is controlled by the firms that profit from the technology, those burdens are likely to be under-priced. 

The better path is to insist that AI policy is shaped by institutions with an explicit mandate to protect long-term human welfare, not to maximize adoption or market share. Governments, regulators, labor organizations, and civil-society groups should be primary rule-makers. Industry can contribute expertise, but it must not be the default agenda-setter. We need transparent, modular, and robust governance structures that can adapt as AI systems evolve. 

AI can be a powerful tool for improving the world, but only if its rules are written by actors whose incentives are aligned with long-term human flourishing, not with maximizing the value of a particular product line. When the referee also owns the game, skepticism is not anti-progress. It is responsibility.