California Vetoes Autopilot DUI Bill Redefining Drive

Have you ever sat in a car that seemed to handle itself and thought, just for a second, that the machine was doing the real work? That question stopped being theoretical in California after a high-profile crash on a major bridge years ago. A driver with a blood alcohol level more than twice the legal limit passed out. The vehicle kept moving with driver-support features engaged. When the person woke up, the defense sounded almost casual: I was not driving. The car was.

That argument is the reason lawmakers tried to rewrite the meaning of drive. On September 27, the governor vetoed the measure. I have followed traffic-law fights for a long time, and this one is messier than a simple yes-or-no on public safety. It sits at the crossing of criminal law, software design, advertising language, and what people think a car can do when they are impaired.

Why California Blocked The Autopilot DUI Rewrite

The bill would have expanded certain driving-under-the-influence, reckless driving, and vehicular manslaughter rules when a vehicle used autonomous or driver-support technology. In plain terms, supervising a car that still needs a person to steer, brake, or accelerate would have counted as driving. That sounds tidy. The veto message said it was not.

The governor said he shared the goal of stopping people from hiding behind partial automation after drinking or using drugs. Then he drew a hard line. As drafted, the language created significant uncertainties in DUI law. Those uncertainties, he warned, could make it harder to hold dangerous drivers accountable, which is the opposite of what supporters wanted.

I share the author’s goal of ensuring that drivers cannot avoid responsibility for impaired driving by engaging partially automated technology.

That sentence is doing a lot of work. It is both a political olive branch and a legal warning. Support the purpose. Reject the drafting. I have found that this pattern shows up whenever legislatures try to freeze fast-moving vehicle software into a criminal statute written for steering wheels and human reflexes.

What The Bill Tried To Change

The core move was definitional. For selected offenses, drive would have included the volitional movement of a vehicle with driver-support features engaged if those features still required a person to supervise and, when needed, steer, brake, or accelerate to keep the trip safe. That is a mouthful. It is also an attempt to close a loophole that some defendants have already tested in public.

The author of the measure, a Democratic assembly member from Santa Cruz, pointed back to that 2018 San Francisco Bay Bridge case. After waking, the driver claimed the car, not the person, was in charge because assistance systems were on. Lawmakers heard a growing pattern: people treat marketing language as a permission slip. If the cabin screen says the system is helping, some assume the human job has ended.

Club organizations that represent large groups of motorists backed the bill. They argued it closed a gap in DUI enforcement and discouraged misuse of systems that are only partly automated. That support mattered. Those groups are usually careful about new criminal language because their members sit on both sides of a traffic stop.

The Drafting Problem That Triggered The Veto

Here is where the veto gets specific, and where I think the governor had a real point even if the politics feel frustrating. The bill used terms that are not aligned with established standards. Driver support features is one example. That phrase may make sense in a press release. It is not a clean match for how regulators, engineers, and courts already classify levels of automation.

Once a statute leans on undefined product language, a roadside stop can turn into a software seminar. An officer would need more than blood-alcohol evidence. Prosecutors might have to prove what the stack of cameras, radar, and control software could and could not do at that exact moment. Courts would then decide whether the person was legally driving before they even reach impairment.

That is not a small delay. DUI cases already depend on timing, observation, and chemical tests. Add a fight over firmware version and feature names, and some cases get slower, some get weaker, and a few collapse. The veto framed that risk as an unintended consequence. Fair enough. Criminal law hates fog.

The Human Story Behind The Legal Fight

People do not usually start the night planning a test case. They drink. They overestimate a system. They trust a brand promise more than the owner manual. Then physics shows up. A bridge. A lane change. A body that slumps. The machine keeps the wheels rolling until it cannot.

The assembly member said these incidents are happening more often because consumers are misled into thinking it is acceptable to operate a semi-autonomous vehicle while impaired as long as the assistance system is engaged. She also argued that advertising language exaggerates what the vehicles can do and leaves people thinking the car is fully self-driving. That last claim is the political flashpoint.

One company and its chief executive have disputed California accusations of misleading advertising about driver-assistance branding. Court filings have argued the statements were not misleading and were protected speech. The bill itself did not name a manufacturer. Everyone in the room knew which systems the public would picture first.

In my experience, that gap between brand poetry and engineering limits is where ordinary people get hurt. A name that sounds like a pilot can feel like a license. It is not. Supervision is still the job. Alcohol and drugs wreck supervision. That is the entire point of DUI law, automation or not.

How Partial Automation Actually Works On The Road

Most systems in passenger cars today are not robot chauffeurs. They keep a lane, hold a gap, and manage speed under conditions the software recognizes. They still want hands near the wheel, eyes on the scene, and a person ready to take over. That readiness is the legal and moral center of this debate.

If you have used these features on a long freeway stretch, you know the strange calm they create. The car tracks paint lines. The cabin gets quiet. Attention drifts. That drift is dangerous even when a driver is sober. Add impairment and the supervision promise becomes fiction.

• Lane keeping still fails in poor markings, construction zones, and sudden geometry changes.
• Adaptive speed control can misread stopped traffic, curves, or unusual vehicles.
• Takeover alerts only help if the person is awake, oriented, and able to act.
• Marketing names can sound complete even when the manual says the driver remains responsible.

None of that requires a conspiracy. It requires ordinary product design plus ordinary human overconfidence. The statute tried to say: if you still have to supervise, you are driving. The veto said: say it with terms courts already understand.

What Police And Prosecutors Would Have Faced

Imagine a night stop. The officer smells alcohol. The occupant is slow. The vehicle was moving with assistance engaged. Under current practice, movement plus control plus impairment is often enough to start a DUI case. Under the proposed wording, someone could argue about whether the software counted as a driver-support feature that required supervision in that exact configuration.

That argument might be weak on the facts. It still costs time. Defense counsel would ask for manuals, logs, and feature lists. Was auto-steer on? Was the driver-monitoring camera working? Did the system demand a hand on the wheel? Which software build was in the car? Those are fair discovery questions in a civil products case. They are heavy cargo in a routine impaired-driving prosecution.

Perhaps the most interesting aspect is that a sloppy definition can help the very people the bill wanted to stop. If the state must first win a technology fight, some impaired occupants get leverage. The governor called that an unintended consequence. I would call it predictable.

Advertising Language And Consumer Confusion

Words sell cars. Words also shape behavior after the second drink. If a system is framed as a copilot, some buyers hear substitute pilot. Regulators have been circling that gap for years. The bill’s author said confusion is worsened when companies use language that stretches capability.

I am not going to pretend every buyer reads the disclaimer in eight-point type. They do not. They remember the demo video. They remember the name on the stalk. They remember a friend who said the car parked itself. Then they make a bad decision on a Friday night.

That does not erase personal responsibility. It does explain why a legislature tried to write the responsibility into the criminal code in a more explicit way. The veto does not bless impairment behind a glowing dashboard. It says rewrite the sentence so police and judges are not decoding product vocabularies before they apply a century of DUI doctrine.

What Existing DUI Law Already Covers

This is the part that gets skipped in social-media summaries. In many cases, a person who is in the driver’s seat, who caused the vehicle to move, and who is impaired can already be charged. The disputed defense is not that alcohol is legal. It is that the human was not driving because software was active.

Courts can reject that defense without a new definition if the facts show the occupant still had a duty to control the vehicle. The bill tried to remove doubt in advance. Removing doubt is good. Doing it with novel labels is risky. That tension is the whole veto.

1. Establish that the person was in a position of control or required supervision.
2. Show the vehicle moved on a public way or in circumstances the statute covers.
3. Prove impairment by observation, testing, or both.
4. Connect the impairment to the risk the law is designed to stop.

Step one is where automation arguments land. If a future bill defines that step with standard automation levels instead of marketing-adjacent phrases, the same policy may survive. The governor invited lawmakers to try again next year. That is not a burial. It is a rewrite request.

Why Next Year’s Version May Look Different

A smarter draft would borrow language that safety agencies and engineers already use. Levels of driving automation exist for a reason. They separate systems that require constant supervision from systems that can perform the dynamic driving task in a defined domain. Criminal statutes do not need to become engineering textbooks. They do need words that map onto those categories without inventing a third dialect.

Lawmakers could also separate two problems that got glued together. One problem is impaired people using assistance features as a crutch. The other is advertising that oversells those features. Mixing both in one criminal definition made the bill feel like a products case wearing a DUI coat. A cleaner bill would keep the coat and leave the products fight for regulators and civil courts.

I would also expect more input from prosecutors who try these cases every week. They know which phrases juries understand. They know which extra elements collapse a file. If the next draft asks them to prove a software taxonomy beyond a reasonable doubt, they will say so earlier.

The Safety Stakes Are Not Abstract

Impaired driving remains one of the most preventable ways to wreck a night, a family, and a roadway. Adding software does not shrink that harm. It can hide it for a few seconds longer. Those seconds matter on a bridge, in a work zone, or in stop-and-go traffic where a system suddenly needs a human who is no longer a human in any useful sense.

Supporters were not inventing a scare story. A person who is blacked out is not supervising anything. If the car is still moving because a feature has not yet given up, the public is exposed. Closing that exposure is a legitimate state interest. The veto did not deny the interest. It denied this wording.

Unfortunately, this bill as drafted introduces significant uncertainties into DUI law, which could have the unintended consequence of making it more difficult to hold dangerous drivers accountable.

Read that again. The fear is not over-criminalization. The fear is under-enforcement caused by fuzzy terms. That is a different critique than the usual culture-war script, and it is why this veto is more interesting than a simple industry win narrative.

Industry, Liability, And The Temptation To Point Fingers

Whenever a crash involves assistance features, three stories compete. The occupant was reckless. The software failed. The advertising confused everyone. All three can be true in different doses. A criminal statute is a blunt tool for that mix.

Companies will keep arguing that drivers are told to stay alert. Safety advocates will keep arguing that names and demos teach the opposite. Legislatures will keep trying to pick a side with a definition. I have watched that loop long enough to know that definitions only work when they are boring. Exciting words make ads. Boring words make enforceable law.

There is also a market angle hiding in the legal file. Driver-assist packages are a selling point. They shape valuations, product roadmaps, and public trust. A criminal rule that turns every impaired-occupant case into a capability hearing would spill into civil discovery and reputation risk. That does not mean the rule is wrong. It means the drafting has to be tight or the spill becomes the story.

Practical Advice For Drivers Right Now

The veto did not create a free pass. If you are impaired, do not get behind the wheel, period. Do not treat lane keeping as a babysitter. Do not treat a traffic jam feature as a nap button. The old rule still fits the new cabin: if the car can move and you are the one who put it in motion, you own the risk.

If you ride with someone who talks as if the car can handle a night out, say something unfashionable and useful. Take the keys. Call a ride. Sleep on a couch. The legal gray zone is not a safety zone. It is a paperwork zone for lawyers after the damage is done.

• Assume every current consumer assistance system still needs a sober supervisor.
• Treat feature names as marketing, not as a legal status.
• If you cannot pass a field test, you cannot supervise software.
• Plan the trip home before the first drink, not after the system chime.

How Courts May Treat These Cases Without The New Statute

Judges already know how to look at control. Who started the trip? Who could have braked? Who was in the seat that the design treats as the fallback? Those questions do not vanish because a camera watched the driver’s eyes or a stack of chips held a lane for a mile.

Defense arguments will continue. Some will fail quickly. A few may find a sympathetic fact pattern: a system that truly required no supervision in that moment, or a vehicle closer to higher automation than the typical passenger car. Most cases will still look like a person who should not have been in that seat.

The absence of the new definition means more judicial discretion and more variation by county. That is sloppy from a policy view. It is also how criminal law often works when statutes lag technology. The next bill can reduce that variation if it stays close to language officers already use in reports.

A Broader Question About Machines And Blame

Every generation gets a new machine that seems to absorb responsibility. Autopilot in aircraft. Cruise control on highways. Now camera-rich driver support in family cars. Each time, people test the edge. Each time, the law eventually says the human remains on the hook until the machine is actually in charge under recognized rules.

We are not at that recognized handoff for most vehicles on California roads. We are in the awkward middle. The middle is where slogans thrive and statutes stumble. I think that is why this veto feels unsatisfying and still coherent. The public wanted a clear no. The governor wanted a clear statute. Those are not the same request.

Policy test for the next draft:
  1. Use terms already known to courts and engineers.
  2. Keep the impaired-occupant duty explicit.
  3. Avoid turning a DUI stop into a software trial.
  4. Leave advertising disputes to the proper forum.

What To Watch In The Coming Session

Look for a revised definition that cites supervision duties without inventing a product category. Look for prosecutor associations to speak earlier. Look for motorist clubs to stay involved, because they supported the last version and will care how the next one treats their members. Look for manufacturers to argue that existing law is enough if manuals already assign responsibility to the human.

Also watch whether the conversation stays on impairment. It should. Lane-keeping while sober and tired is a different policy pile. Reckless speed with assistance on is another. This bill was about alcohol and drugs plus partial automation. Keeping that focus will make a rewrite more likely to pass.

If lawmakers wander into a general referendum on branded assistance systems, the file gets heavier and the veto risk rises again. Tight bills survive. Culture-war bills get quotes and then stall. That is not cynicism. That is calendar math.

The Bottom Line After The Veto

California still wants impaired people off the road. The governor said as much. Motorist groups still want the loophole closed. The author still has a vivid case from a bridge that made the abstract concrete. None of that disappeared on September 27. What disappeared was a particular paragraph that tried to redefine driving with language the executive branch called too uncertain.

So the gray zone remains, at least on paper. In real life, the safe choice is unchanged and almost boring. If you cannot drive, do not start the trip and do not outsource judgment to a feature that still needs you. Software can hold a lane. It cannot hold your judgment.

Next year will tell us whether the legislature can write that idea in words a patrol officer can apply at midnight without calling an engineer. Until then, the veto stands as a reminder that good intentions do not enforce themselves. Clear statutes do. And clear statutes, in this corner of the law, have to be a little less poetic than the names on the dashboard.