Eighty-five percent. That’s the share of trips a proposed New Jersey bill circulated by Uber lobbyists would require human drivers to handle, according to reporting on the fight between Uber and Waymo. Not a technical benchmark. Not a safety threshold from an engineering review. A number written into draft legislation.
Which tells you something important about how self-driving cars will actually arrive in your city: the deciding factor may be less about sensors and software than about who has more people in the building when the bill gets marked up.
Waymo, the autonomous vehicle company owned by Alphabet, has doubled what it spends on lobbying. That’s a 100% jump in the budget line dedicated to shaping U.S. rules for robotaxi operations. And it’s happening in direct competition with Uber, a company that has spent more than a decade getting very, very good at exactly this.
Why an AI Explainer Cares About Lobbying Budgets
I usually write about how AI agents work. How they perceive, decide, and act. A robotaxi is the most physical AI agent most people will ever meet: it senses the road, plans a route, makes thousands of small decisions, and moves two tons of metal through a neighborhood where children live.
So when I look at an AI agent, I tend to ask technical questions. Can it handle a four-way stop with a confused cyclist? What happens in heavy rain? How does it behave when a police officer waves it through a red light?
Those questions matter enormously. But they’re not the questions that determine whether the agent gets to operate. That gets decided in state capitols, city councils, and regulatory agencies. A robotaxi can be technically ready and legally banned. It can also be legally permitted and practically unwelcome, buried under permit requirements and reporting rules that make the economics impossible.
Waymo doubling its lobbying spend is an admission that the hard part of deploying an AI agent at scale isn’t only the model. It’s the rulebook.
Two Companies, Two Very Different Bets
What makes this rivalry interesting is that Uber and Waymo want different things from regulators, because their businesses depend on different futures.
- Waymo builds the driving system itself. Its value grows as human drivers become optional. Rules that permit driverless operation across more cities and more conditions are rules that expand its market.
- Uber runs the marketplace that connects riders with drivers. Its network is enormous and human. Rules that keep humans in the loop protect a business model that already works at massive scale.
Reporting on the D.C. fight describes Uber working alongside labor unions, an alliance that makes strategic sense from both sides. Drivers want their jobs protected. Uber wants a transition it controls rather than one that arrives on Waymo’s schedule.
That’s not a villain story. It’s two companies with different economics arguing for different rules, which is what lobbying has always been. What’s new is that the subject of the argument is an AI system rather than a tax code or a zoning variance.
What This Means for Regular People
If you’ve been waiting for someone to tell you whether robotaxis are safe enough, here’s the uncomfortable part: the answer you eventually get may be shaped as much by advocacy as by evidence. Rules written under pressure from competing interests can end up somewhere reasonable. They can also end up arbitrary.
An 85% human-driver requirement isn’t derived from a crash study. It’s a negotiating position. It might be a sensible brake on a fast-moving technology, or a moat dressed up as caution. Both readings are available, which is precisely why the number is useful to whoever proposed it.
A few things worth watching as this plays out:
- Where the rules get written. State legislatures and city councils are doing the heavy lifting here, not a single federal agency. That means outcomes will vary wildly by geography.
- Who shows up to testify. Safety researchers, disability advocates, transit riders, and drivers all have stakes. Whether they get airtime alongside well-funded lobbyists shapes the result.
- Whether performance data becomes public. The strongest argument against lobbying deciding this is real, verifiable operating data that anyone can examine.
The Part Nobody Puts in a Product Demo
We tend to talk about AI progress as a series of technical milestones. Better perception. Better planning. Fewer interventions. Those milestones are real, and Waymo has racked up plenty of them.
But a doubled lobbying budget is also a milestone, just a less photogenic one. It marks the moment a technology stops being a research problem and becomes a political one. Every AI agent that touches the physical world, or a regulated industry, or someone’s job, will eventually reach this stage.
Robotaxis got here first. Understanding how their rules get made is good practice for everything that follows.
🕒 Published: