OpenAI just learned that putting AI regulation on a ballot doesn't mean everyone thinks you're the good guy.
The Summary
- A coalition of child safety and digital rights groups is demanding OpenAI withdraw its California ballot measure, arguing it would actually weaken protections for kids while shielding AI companies from liability.
- The measure, backed by OpenAI's lobbying arm, would preempt stronger future legislation and lock California into a framework the company helped design.
- This is regulatory capture in real time: the regulated writing the rules, then asking voters to make them permanent.
The Signal
OpenAI has been funding a California ballot initiative that positions itself as "AI safety legislation" focused on protecting children online. Sounds good until you read what advocacy groups actually see in the fine print. The coalition, which includes the Tech Transparency Project and Accountable Tech, says the measure would establish weak baseline protections while blocking California from passing stronger rules later. It's the legislative equivalent of a chess move: sacrifice a pawn now to control the board forever.
The playbook here is familiar. Tech companies have been running ballot initiatives for years to bypass legislatures they can't fully control. Uber and Lyft did it with Prop 22 to reclassify drivers. Now OpenAI is trying the same move with AI governance. The difference is timing. We're still in the early days of understanding what AI agents actually do to kids, to labor markets, to information ecosystems. Locking in rules now, before we even know what problems we're solving, is either naive or strategic. The coalition is betting on strategic.
What makes this particularly sharp is that OpenAI has positioned itself as the responsible AI company, the one that paused to think about safety while others raced ahead. But when you fund a ballot measure that limits your own legal accountability while freezing the regulatory landscape, you're not building safety infrastructure. You're building a moat. The concern isn't just what the measure does today. It's what it prevents tomorrow when California legislators wake up to harms we haven't named yet and find their hands tied by a voter-approved initiative written by the company they're trying to regulate.
The Implication
If you're building in AI, watch how this plays out. It's a test case for whether companies can successfully use direct democracy to bypass legislative oversight. If it works in California, expect it everywhere. If you're a voter in California, read the actual ballot language when it shows up, not the marketing campaign. The question isn't whether you trust OpenAI today. It's whether you want to lock in rules before we know what we're regulating.
Source: Decrypt