Auto insurance exists to price risk. That risk has always been human risk. Drivers get distracted, tired, angry, careless, and drunk. They misjudge distance and speed. They fail to react in time. For decades, insurers have tried to manage this reality indirectly, by sorting drivers into demographic buckets, pricing by age, zip code, driving history, and crude proxies for behavior. The system has never been elegant, only tolerable. Full Self-Driving changes this arrangement at its foundation.
The core claim is simple. If driving becomes dramatically safer when software is in control, insurance must become dramatically cheaper. Not in theory, but in practice. We are now seeing the first concrete evidence that this shift has begun.
Next month, Lemonade will launch a new form of auto insurance in Arizona, with Oregon to follow. It is the first insurance product designed explicitly around Tesla’s Full Self-Driving system. The policy integrates directly with Tesla vehicle telemetry. It observes, mile by mile, whether the car is being driven by a human or by FSD. Miles driven under FSD are priced at half the rate of human-driven miles.
This is not a marketing gimmick. It is an actuarial judgment. Lemonade’s pricing reflects a belief that FSD miles carry substantially lower accident risk than human-driven miles. The insurer is not guessing. It is ingesting real-time data from the vehicle itself, including software version, sensor usage, and engagement status. Risk is no longer inferred. It is observed.
Why does this matter? Because insurance pricing follows loss frequency. When accidents fall, claims fall. When claims fall, premiums must follow. Any insurer that fails to pass those savings to consumers will be undercut by one that does.
To see why this is inevitable, consider current insurance costs. The average U.S. driver now pays roughly $2,000 per year for auto insurance, with many paying far more. In high-cost states, annual premiums routinely exceed $3,000. These prices reflect a grim reality. Crashes are common. Injuries are expensive. Litigation is endemic. Human drivers impose enormous costs on one another.
Now consider what happens when half or more of those miles are driven by a system that crashes far less often.
Tesla’s own safety data already points in this direction. In Q3 2025, Tesla reported one crash per 6.36 million miles driven with Autopilot or supervised FSD engaged. The U.S. national average sits closer to one crash per 0.7 million miles. Even allowing for favorable driving conditions, the gap is enormous. It suggests a risk reduction on the order of multiple factors, not marginal improvements.
Lemonade’s decision to discount FSD miles by 50% is therefore conservative. If the true risk reduction were fully priced, premiums could fall far more. The insurer is starting cautiously because current FSD is supervised. The human driver must remain attentive. Responsibility is shared. As long as that is true, some human risk remains in the system.
But even this partial autonomy produces meaningful savings. If a driver uses FSD for half their miles, and those miles are priced at half the rate, total premiums fall by roughly 25%. Increase FSD usage to 80%, and savings approach 40%. As the system improves, and as confidence grows, further reductions follow naturally.
Independent evidence reinforces this logic. Waymo’s fully driverless vehicles offer a glimpse of what happens when human error is removed entirely. Over millions of miles, Waymo’s autonomous fleet has demonstrated substantially lower crash rates than human-driven vehicles in comparable environments. Injury-causing crashes drop even more sharply. These are not simulations. They are real-world results on public roads.
The explanation is not mysterious. Human error causes roughly 94% of serious crashes. Eliminate the human, and you eliminate the dominant failure mode. An AI driver does not text. It does not fall asleep. It does not rage. It sees in all directions at once and reacts in milliseconds. It obeys traffic laws by default. Even an imperfect system with these properties will outperform the median human driver.
Insurers know this. Reinsurers know this. Swiss Re has already reported massive reductions in bodily injury and property damage claims for autonomous fleets compared to human-driven cars. Financial analysts have taken notice as well. Goldman Sachs projects that auto insurance costs could be cut roughly in half over the coming decades as autonomy spreads. EY projects similar declines.
What Lemonade has done is translate these abstract forecasts into an operational product. It prices risk by mode, not by person. When the AI drives, the AI’s safety record governs the premium. When the human drives, human risk returns.
This is a profound shift. Traditional insurance treats the driver as the locus of risk. Age, gender, credit score, and past violations dominate pricing. Mode-based insurance treats control as the locus of risk. Who is driving matters less than what is driving.
This approach has several consequences. Younger drivers benefit immediately when FSD is engaged. High-risk drivers can neutralize much of their risk by ceding control to the system. Safe drivers are no longer cross-subsidizing reckless ones when automation is active. Risk is priced where it actually resides.
Tesla has already gestured in this direction with its own insurance product, offering modest discounts when FSD usage exceeds certain thresholds. Lemonade goes further by pricing each mile independently. The difference reflects institutional position. Tesla must tread carefully under regulatory scrutiny. Lemonade, as a third-party insurer, can move faster.
The implications extend beyond pricing. As autonomy improves, traditional rating factors will erode. If a vehicle truly drives itself, driver age becomes irrelevant. Driving history becomes irrelevant. The AI driver is the same for everyone. In such a world, a 20-year-old and a 50-year-old pay the same rate when the car is in control.
This is not speculative philosophy. It is a straightforward consequence of assigning liability to the agent that acts. When software drives, software risk dominates. When humans intervene, human risk returns.
Some worry that insurers will resist this transition to protect premium volumes. That resistance cannot last. Insurance is a competitive market. Firms that overprice low-risk behavior lose customers. Firms that underprice high-risk behavior go bankrupt. Lemonade’s product will either validate its assumptions or force rapid correction. Either way, the signal is sent.
There is also a broader feedback loop at work. Cheaper insurance incentivizes greater use of FSD. Greater use generates more safety data. Better data supports lower premiums. Lower premiums accelerate adoption. Each step reinforces the next.
Over time, the economics become unavoidable. If autonomous driving reduces crashes by even 50%, insurance premiums must fall by roughly the same amount. If reductions approach the levels seen in fully driverless fleets, premiums collapse further. Insurance does not disappear, but it shrinks. Liability migrates. The human-centered model dissolves.
Lemonade’s FSD insurance marks the beginning of this process. It is the first clear instance of insurance pricing catching up to technological reality. It treats autonomy not as a novelty, but as a measurable reduction in risk. Others will follow, because they must.
Auto insurance has always been a tax on human fallibility. Full Self-Driving offers a way to reduce that tax dramatically. The numbers already point in one direction. As AI replaces human judgment behind the wheel, safer roads and cheaper insurance will not be optional outcomes. They will be the natural result of aligning price with risk.
If you enjoy my work, please subscribe: https://x.com/amuse.
Sponsored by the John Milton Freedom Foundation, a nonprofit dedicated to helping independent journalists overcome formidable challenges in today’s media landscape and bring crucial stories to you.
READ NEXT: Bone-Chilling Legal News Rocks The Clintons
See Auto Co. Tensor as model for SDA
Self Drive markets:
Seniors who cant drive due to Medical issues
Those disabled but can move
alone
$$$$$$$
See Tensor.com
“Next month, Lemonade will launch a new form of auto insurance in Arizona, with Oregon to follow. It is the first insurance product designed explicitly around Tesla’s Full Self-Driving system. The policy integrates directly with Tesla vehicle telemetry. It observes, mile by mile, whether the car is being driven by a human or by FSD. Miles driven under FSD are priced at half the rate of human-driven miles.”
“Tesla’s own safety data already points in this direction. In Q3 2025, Tesla reported one crash per 6.36 million miles driven with Autopilot or supervised FSD engaged. The U.S. national average sits closer to one crash per 0.7 million miles. Even allowing for favorable driving conditions, the gap is enormous. It suggests a risk reduction on the order of multiple factors, not marginal improvements.
Lemonade’s decision to discount FSD miles by 50% is therefore conservative.”
In other words, the data above suggests insurance companies are setting themselves up to make massive profits by only discounting a small portion of what they are actually saving.
I just don’t have enough faith in the system yet. Being 70, right now, I know my reflexes aren’t they used to be, but in my years of driving, I’ve been in an accident with another car 1 time. I was stopped at a 4 way stop, and got rear ended by a 14 yr. old girl. As of right now, I feel I can compete with the computers. I hope I can for many more too!