AUTOMOTIVE EXCESS MEETS REGULATION — Tesla’s Got the Wrong Horse for this Rodeo

Tesla Brought Mad Max to the Wrong Continent

A single Swedish regulator’s objection to a driving mode is now the reason Tesla’s entire European rollout might slip into 2027. It was the right call.

On 15 April, a Swedish Transport Agency investigator named Hans Nordin sat down to write an email and discovered something that genuinely surprised him. Tesla, the company asking his government to approve its driving software for use on Swedish roads, had built that software to let drivers configure it to break the speed limit. Not accidentally. Not as a bug awaiting a patch. As a named, marketed feature, sitting in a settings menu, waiting for someone to turn it on.

Nordin is not, by training or temperament, a man given to surprise. His job is to read regulatory submissions and find the part that doesn’t hold up. He found it almost immediately, five days after the Netherlands had become the first EU country to approve the system.

On 30 June, the European Union’s Technical Committee on Motor Vehicles meets in Brussels to continue discussing whether to approve Tesla’s Full Self-Driving system across all twenty-seven member states. No vote is scheduled — the published agenda lists the item as a continuation of talks, fifty minutes set aside for a question that has already taken months to not resolve. Sweden has already told the committee, in writing, what its vote will be whenever one finally happens. The recommendation, dated 30 April, is unambiguous: reject it, unless Tesla removes the feature that let Nordin’s surprise happen in the first place.

This is not, on its face, a large dispute. The feature is called Speed Offset. It lets an owner tell the car how many kilometres per hour above the posted limit it is permitted to drive — a kind of permission slip for the accelerator, written by the driver and executed by the software. Tesla built it because, by its own logic, real drivers don’t drive at the speed limit. They drive a bit over it, most of the time, and a car that refuses to do the same will feel stiff, foreign, untrustworthy. Match the traffic, not the sign. It is a design philosophy with an almost coherent internal case.

This publication is reader-funded — no ads, no sponsors, no brand deals. If you want more long-form writing like this to exist: **☕ Buy me a coffee** — it funds the next one directly.

Sweden’s regulators are not interested in the internal case. Their objection, in the words of the letter sent to the committee, is that “allowing automated systems to systematically exceed legal speed limits risks undermining both the legal framework and the expected safety benefits of vehicle automation.” That sentence is doing more work than it first appears to. It is not a complaint about a feature. It is a complaint about a premise — the premise that a machine should be permitted the same casual relationship with the law that a human driver has always quietly enjoyed.

Finland and Norway have said similar things, more bluntly. Norway’s Stein-Helge Mundal, asked by Tesla owners to wave the system through, told them Tesla’s own safety figures “are self-produced,” and that he could find no way to correlate them with his country’s actual accident statistics.

That is as close as you will ever hear a Nordic transport official get to calling a number a complete fiction without using the word.

And here is where the dispute stops being about one feature and starts being about a pattern. Because the same week Sweden filed its objection, reporting drawing on Reuters correspondence emerged showing Tesla had handed European regulators a second number worth examining — and that number doesn’t survive contact with anyone who checks it.

The claim, presented in a slide deck Tesla’s policy manager emailed Swedish officials within days of the company’s first European approval, was that FSD-equipped cars travel more than seven, even ten, times farther between crashes than the average American driver. It also claimed that adopting the system widely could have saved 32,000 lives. Both are striking statistics. Both are also, according to the eleven independent traffic-safety researchers Reuters asked to review them, unstable ones — ten of the eleven called the methodology misleading marketing rather than a serious safety study. More fiction.

Tesla’s comparison counts only crashes severe enough to deploy an airbag on its own side of the ledger, then measures that narrow figure against the far wider category of all towed-vehicle crashes in US national data — categories that aren’t counting the same thing. Marco Benedetti redid the comparison on a like-for-like basis, airbag deployments against airbag deployments on both sides. He is an assistant research scientist at the University of Michigan Transportation Research Institute and a former NHTSA statistician, and his correction collapsed the claimed advantage from ten times to roughly three. Even that residual figure doesn’t survive a second look, Benedetti noted: Tesla’s fleet averages 4.1 years old, against a thirteen-year-old national average. The comparison set newer cars with modern automatic braking and lane-departure systems against a baseline stuffed with vehicles that predate them entirely. As Carnegie Mellon’s Phil Koopman put it, that’s rather like boasting that your jet outruns somebody’s WWII bomber.

Catching that kind of error doesn’t require institutional power. It requires someone willing to ask what the numbers were actually measuring, which is precisely the question that Article 39 of the EU’s vehicle regulation architecture was never built to ask. A separate Swedish investigator, Anders Eriksson, told Reuters that regulators “look beyond headline figures” and would not base an assessment “solely on aggregated safety claims” — a careful sentence that stops short of saying whether Tesla had offered much else. Under Article 39, once one member state grants a system provisional approval, others can recognise it without running their own tests — useful, efficient machinery for verifying brake performance or headlight geometry, where the physics is shared and the answer is singular. It is considerably less useful for evaluating whether a manufacturer’s lobbying slide deck survives a statistics seminar. Lithuania, Estonia, Denmark and Belgium have all recognised the Dutch regulator’s original approval on exactly that basis — without independently revisiting whether the data behind it was sound. That was exceedingly unwise.

There is a steelman here, and it deserves a fair run before it gets dismantled. The case for Speed Offset is not stupid. Traffic, as anyone who drives in it knows, does not move at the posted limit; it moves at the limit plus a few, and a vehicle that rigidly refuses to join that flow becomes the obstacle, not the example. There is a genuine, defensible argument that matching real traffic behaviour is safer than enforcing a number nobody else on the road is observing. Tesla is not wrong that human driving has always included this unwritten margin.

But are we talking road safety? Or simple speeder’s annoyance?

This particular steelman has a structural problem, and it’s the same one that undid the safety statistics. Tesla isn’t asking regulators to evaluate whether human-like flexibility is safe. It’s asking them to certify, in regulatory text, that a machine may be configured to break the law as a documented setting — and then defending that ask with numbers that don’t hold up when somebody actually runs them twice. A regulatory culture that treats the speed limit as load-bearing law, not a polite suggestion, was never going to find “but everyone does it anyway” a persuasive argument, however true it is on the motorway. And a regulatory culture whose own staff recently watched a vendor’s safety claim fall apart under a basic severity-threshold check is not in a generous mood to extend trust on the next claim either.

It is worth a short detour here, because the pattern predates Sweden’s formal objection by several months. In the United States, Tesla’s domestic FSD product offers drivers a menu of named personalities — Chill, Hurry, and, reintroduced to predictable controversy in October 2025, Mad Max, a mode whose entire marketed identity is aggression: tighter gaps, faster lane changes, less patience for the posted number. It is surely not just European regulators who see Mad Max Mode as not a good thing on suburban roads?

Tesla’s European build deceptively carries none of that branding. It offers Speed Offset and a more diplomatically named Contextual Max Speed instead. The substance survived the journey across the Atlantic; only the marketing changed costume. That a company can correctly anticipate that European regulators would baulk at “Mad Max” while still trying to sell them the underlying capability, and call it Contextual, suggests Tesla understood exactly how this would read — and decided to test how far the substance could travel once the name was quieter. That is to say, how far their deception would travel.

The European Transport Safety Council, watching all of this from the outside, has put the obvious remedy on the table: if Tesla’s safety claims are real, hand the underlying data to an independent university researcher, let it be checked properly, and then come back and talk. Tesla has not taken up the offer. The Dutch regulator that granted the original approval insists it relied on its own testing, not on marketing claims — but has not said whether it specifically interrogated the validity of the American figures Tesla submitted before that approval went on to cascade across four more countries.

The arithmetic of 30 June is not complicated, even if the politics are. A qualified majority — fifteen of twenty-seven states, representing sixty-five per cent of the EU’s population — is required for bloc-wide approval, which means the vote effectively runs through Germany, France or Italy, none of which has yet moved. Sweden’s formal recommendation does not by itself sink that count. What it does is tell every other undecided capital that opposing Tesla on a clearly stated principle costs nothing politically — which is a more dangerous thing for Tesla’s timeline than the vote itself.

The committee is not expected to resolve anything on the 30th. It will simply sit in a room in Brussels and listen to representatives from a continent that builds its trust in institutions slowly, and rarely extends it twice to the same applicant for the same reason. Let’s see if Mad Max and the Rule of Law can be found compatible.

Four publications. Pick the ones that fit.

***The Strategic Edge* — Geopolitics, war economy, trade, and macroeconomics. Numbers that explain what’s actually happening.

***The Civic Ledger* — Policy, politics, culture, and society. Charts and cost curves you can use at work and in arguments.

***Torque & Theory* — EV, hybrid, and ICE: the maths, the documents, and how to save money on the decisions that matter.

*The Permanent Secretary — A weekly memorandum on politics, diplomacy, and the management of world affairs. From the office of the Permanent Secretary.*

Follow one. Follow all four. Either way, you’re reading work that isn’t written for the algorithm