Tesla Autonomy & Safety

Claim: Tesla’s Full Self-Driving and robotaxi are vaporware that will never ship.

Reality: Tesla launched a paid Robotaxi service in Austin in June 2025 and began expanding it to more cities, and started production of the purpose-built Cybercab. FSD (Supervised) is used across millions of cars and billions of miles, improving with each release. This is the most legitimately contested item on the site: Musk has repeatedly missed his own self-imposed deadlines (he's promised "robotaxis next year" since around 2019), and fully unsupervised driving for customer cars is not yet delivered. So skepticism about timelines is earned. But "vaporware" means a product that doesn't exist — and a real, paid, on-road robotaxi service plus a shipping autonomy system is the opposite of vaporware. The honest framing is "later than promised," not "never real."

Claim: Cars on Tesla Autopilot crash far more often than normal cars.

Reality: Tesla's published Vehicle Safety Report consistently records fewer crashes per million miles when Autopilot is engaged than when it is not, and far fewer than the US fleet average. Independent researchers rightly note the comparison is imperfect — Autopilot is used mostly on highways, which are inherently safer per mile — so the numbers should be read with that caveat, and regulators continue to scrutinise specific crashes. But the strong claim that Autopilot is "far more dangerous" than human driving is not supported by the available aggregate data; if anything the data points the other way. The honest framing is that the technology is a driver-assistance aid that reduces crashes when used as instructed, while misuse — treating it as hands-off autonomy — is the genuine danger that high-profile crashes usually involve. NHTSA's scrutiny is appropriate and Tesla has shipped over-the-air updates strengthening driver-attention monitoring in response. "A tool that is safer on average but dangerous if abused" is a very different and more accurate claim than "far more dangerous than a human."

Claim: Tesla sells Autopilot as a fully self-driving system that needs no driver.

Reality: Tesla's owner documentation, in-car prompts and purchase flow state that Autopilot and "Full Self-Driving (Supervised)" require an attentive driver with hands ready on the wheel at all times. The naming has been genuinely and fairly criticised by regulators as potentially confusing — that critique has merit. But the official instruction to drivers is supervision, not hands-off autonomy, and the car actively monitors driver attention. The gap is between marketing branding and legal instruction, not a claim that the car drives itself unattended. Tesla also renamed the package to "Full Self-Driving (Supervised)" in 2024 — an explicit acknowledgement of the criticism and a move toward clearer language. The distinction matters because Tesla's genuinely uncrewed system runs under a separate, geofenced Robotaxi program with no driver at all; the consumer car you buy is still, by Tesla's own instructions, a supervised assist that keeps the human responsible.

Claim: The $243 million Benavides wrongful-death verdict proves Tesla Autopilot is dangerous and defective.

Reality: In August 2025 a Miami federal jury found Tesla partly liable for a 2019 crash that killed Naibel Benavides, awarding about $243 million (mostly punitive). It was the first US verdict tying a death to Autopilot, and in February 2026 the judge declined to toss it; Tesla is appealing. That is a real, serious case. But one verdict over a single 2019 incident doesn't establish that the system is broadly unsafe. The driver, George McGee, was reaching for a dropped phone and pressing the accelerator — overriding the car — when it ran a stop sign at ~60 mph; the jury still assigned him the majority of fault. Across the fleet, Tesla's Vehicle Safety Reports show roughly one crash per 6+ million miles with Autopilot engaged versus a US average near 700,000 miles. A tragic misuse case on appeal is not proof of a defective product.

Claim: Reuters proved Tesla fabricated its Full Self-Driving safety statistics.

Reality: A May 2026 Reuters investigation made a genuine, fair criticism: Tesla's headline safety numbers rely on a debatable methodology. Tesla compares airbag-deployment crashes in Autopilot-equipped cars against a broad federal crash rate that includes less-severe, tow-only incidents, and against an older average US vehicle — comparisons most researchers Reuters consulted called misleading. Tesla also doesn't publish the underlying data or peer-reviewed studies, a legitimate transparency gap. But "fabricated" overstates it. Reuters did not show Tesla invented numbers; the dispute is about apples-to-oranges comparisons and transparency, not fraud. Autopilot is also used disproportionately on highways — the safest roads — which cuts both ways. The honest takeaway: Tesla's marketing framing is contestable and it should release raw data; that is different from proof of fabrication.

Claim: Tesla's FSD runs red lights and is so dangerous the government is investigating it.

Reality: There is a real investigation: NHTSA opened a preliminary evaluation in October 2025 covering roughly 2.9 million Teslas after dozens of reports that FSD ran red lights, crossed into oncoming lanes or committed other violations, including some intersection crashes with injuries. By early 2026 the agency had upgraded its review. Those are legitimate concerns worth tracking. But a preliminary evaluation is a routine information-gathering step, not a finding of a defect or a recall — NHTSA opens many that close without action. The reported violations number in the dozens against billions of FSD miles driven, and Tesla ships frequent over-the-air updates targeting exactly these edge cases. The honest position: real edge-case failures exist and the probe is justified, but "FSD runs red lights" as a blanket description of normal operation is not supported by the overall record.

Claim: Musk is a serial over-promiser whose companies never actually ship what he announces.

Reality: The honest half is true: Musk routinely misses his own deadlines. He predicted a million robotaxis by 2020, full Self-Driving "next year" for years running, Starship to orbit in 2022 and a crewed Mars mission by 2024 — none on schedule. That's a real, documented pattern of optimistic timelines, and it's fair to discount his dates. But "never ship" is false. The products keep arriving, just late: SpaceX landed and reflew orbital boosters hundreds of times, Starlink serves millions, the Model Y became one of the best-selling cars on Earth, Tesla launched a paid robotaxi service in Austin in 2025, and Neuralink implanted its device in human patients. The pattern is "aggressive deadline, slips, but delivers a real product," not vaporware. The fair verdict is mostly false: criticize the timelines all you want — they routinely slip — but the claim that the things he promises never materialize is contradicted by a long list of shipped, working products.

Claim: Tesla’s “phantom braking” is a dangerous defect that proves Autopilot is broken.

Reality: The honest part: phantom braking — where the car brakes for a hazard that isn’t there — is real, generated hundreds of NHTSA complaints, and prompted a federal investigation. That’s a legitimate issue and Tesla shouldn’t be let off the hook for it. What’s misleading is the leap from “annoying false positive” to “dangerous defect that proves the system is broken.” Phantom braking is a tuning problem common to camera-and-radar driver-assistance systems across the industry, not unique to Tesla, and it errs toward caution (braking) rather than failing to brake. Tesla has reduced it over successive over-the-air updates, including the move to the vision-based and later end-to-end FSD stacks, precisely because the software can be fixed fleet-wide without a workshop visit. So criticise the false positives — they’re real and irritating — but “proves Autopilot is broken” overstates a calibration issue that has measurably improved.

Claim: Tesla had to recall 2 million cars because Autopilot is unsafe.

Reality: The headline is technically accurate and deeply misleading at the same time. In December 2023 Tesla agreed with NHTSA to address Autopilot driver-monitoring concerns across roughly 2 million vehicles, and US regulators legally classify any such fix as a “recall.” But unlike a traditional recall, no car went to a workshop and nothing physical was replaced: Tesla pushed an over-the-air software update that strengthened driver-attention warnings. The word “recall” conjures images of millions of defective cars hauled in for repair, which simply didn’t happen — it was a remote software change overnight. The underlying issue (making sure drivers stay attentive while using Autopilot) is legitimate, but “recalled 2 million cars because Autopilot is unsafe” weaponises a regulatory label to imply a mass mechanical defect that the facts don’t support.

Claim: Tesla’s Robotaxi is vaporware — it’s just a handful of cars and a safety driver.

Reality: The honest part: Tesla's driverless Robotaxi fleet is still modest in scale and trails Waymo badly on total paid trips, so anyone calling it a finished, mass-scale service is overselling it. But "vaporware" is flatly wrong. From January 2026 Tesla began running Austin Robotaxis with no safety monitor in the car — genuinely driverless, paying-passenger rides, first on a limited share of the fleet and then growing — and by late 2026 the service operated unsupervised across six US cities (Austin, Dallas and Houston in Texas; Miami, Orlando and Tampa in Florida), had passed a million unsupervised miles, and had begun rolling out the purpose-built Cybercab. "Vaporware" means a product promised and never shipped; a driverless car you can hail and ride across six cities is, by definition, shipped. The fair criticism is about scale and pace, not existence — and Tesla's camera-only approach, if it generalises, is far cheaper to expand than lidar-mapped rivals because the hardware already ships on millions of cars.

Claim: Musk’s $1 trillion Tesla pay package proves pure greed.

Reality: The "$1 trillion" headline is real but the "pure greed" reading is misleading because it omits how the package works. It is 100% performance-based: Musk receives nothing unless Tesla clears a ladder of twelve extreme milestones — pushing the market capitalisation toward $8.5 trillion, hitting enormous profit targets, and delivering 20 million vehicles plus millions of Robotaxis and Optimus robots. If those aren't met, the payout is zero, and shareholders only "pay" by first becoming dramatically wealthier. Tesla's own shareholders — the people whose money it is — approved it with roughly 75% support, explicitly to keep the founder focused on the company through its riskiest AI and robotics bets. One can reasonably argue any pay that large is excessive or that the targets could still be gamed, and proxy advisers did object. But framing a zero-unless-you-create-trillions-in-value structure as straightforward greed ignores that he is paid only if everyone else wins first.

Claim: Musk’s politics destroyed Tesla — the brand is toxic and sales are collapsing.

Reality: The honest half is very real and shouldn't be minimised: Tesla's brand took a genuine hit in 2025. European sales plunged sharply — down roughly 49% year-on-year in some months — brand-favourability surveys fell, "Tesla Takedown" protests and showroom vandalism spread, and US deliveries slipped. It is fair to say Musk's polarising politics repelled a slice of Tesla's traditionally progressive buyer base, and that is a self-inflicted wound. But "destroyed Tesla" overstates it, because the sales dip has several causes beyond politics: the best-selling Model Y went through a global changeover that left a production gap, the lineup aged against a wave of new EV competition (BYD especially), price cuts pulled demand forward, and EV subsidies shifted. Tesla remained profitable, kept the largest US EV market share, and its stock recovered later in 2025 as attention turned to robotaxi and energy. So the fair verdict is that the claim is misleading: the political brand damage is real and worth conceding — but a company still building ~1.6 million cars a year and leading autonomy and grid storage has not been "destroyed."

Claim: Tesla illegally lies about "Autopilot" and "Full Self-Driving" in its advertising.

Reality: Regulators have in fact found Tesla's driver-assistance marketing misleading, so this is not a baseless smear. The California DMV brought a case arguing that the names "Autopilot" and "Full Self-Driving" and Tesla's related advertising overstated what the cars can do, and in December 2025 a California administrative judge ruled that Tesla had deceptively marketed the technology and ordered it to fix the marketing (Tesla then sued the DMV to reverse the ruling, and the DMV ultimately declined to suspend Tesla's sales license). This echoes an earlier 2020 German court ruling against Tesla's Autopilot advertising. The defense is not that the names are perfect — reasonable people think "Full Self-Driving" oversells a system that legally requires an attentive driver — but that Tesla does disclose the supervision requirement in its materials and in the car, that the underlying technology is genuinely among the most capable driver-assistance systems sold, and that a naming dispute is a marketing-truth question, not proof the cars are unsafe or that owners were sold a product that doesn't exist. Verdict: mostly true — a regulator did rule Tesla's Autopilot/FSD marketing deceptive and in breach of state law, so the core of the criticism holds; "illegally lies" slightly overstates a civil ruling Tesla is still fighting in court.

Claim: Musk is a serial over-promiser who never actually delivers what he announces.

Reality: The pattern is real and worth conceding: Musk routinely sets deadlines his companies miss. Year-end 2025 roundups catalogued a long list of slipped targets — most prominently Tesla robotaxis "everywhere" and big FSD and Optimus milestones that arrived late or not at all — and he has a documented history of aggressive timelines, from "Mars by" dates to "full self-driving next year" repeated for years. Anyone taking his calendar predictions literally will be disappointed, and that is a fair criticism. What the "never delivers" version gets wrong is the outcome, as opposed to the schedule. The same person who is chronically late has also, eventually, done the things skeptics called impossible: landing and reusing orbital rockets, catching a Super Heavy booster with the launch tower, building the best-selling car on the planet, deploying the largest satellite-internet constellation in history, and launching an actual paid robotaxi service in Austin in 2025 — late and limited, but real. The honest verdict is that Musk's timelines are unreliable while his eventual delivery on the core goal is far better than "never." Verdict: mostly false — chronic over-promising on dates is real; "never delivers" is contradicted by a long record of eventually shipping.

Claim: Elon Musk chronically over-promises and misses his own deadlines.

Reality: This one is largely true, and conceding it is more honest than spinning it. Musk's timeline record is full of misses: he said a Tesla would drive itself coast-to-coast with no human intervention by the end of 2017 (it still has not, on the terms promised), told an April 2019 audience he was "very confident" there would be over a million robotaxis on the road in 2020 (there were none), repeated that full self-driving was roughly a year or two away in 2016, 2018, 2019 and most years since, suggested crewed Mars landings by the mid-2020s, and unveiled the Cybertruck for 2021 production that slipped to late 2023. The "$35,000 Model 3" he promised for years was offered only briefly before being pulled. As a predictor of dates, he is routinely, sometimes wildly, optimistic. What keeps this at "mostly-true" rather than a flat indictment is that the goals themselves — not just the dates — have an unusual habit of eventually arriving. SpaceX did land and reuse orbital rockets, did fly NASA astronauts, did catch a booster with the launch tower, and did build Starlink into a global network; Tesla did ship the Cybertruck and did put genuinely driverless robotaxis (no safety monitor in the car) onto public roads from 2026, scaling to six US cities by September 2026 — each late, and the robotaxi still far smaller than the fleet he promised, but real. The fair criticism is that Musk treats aggressive deadlines as motivational targets rather than commitments, which misleads customers and investors who take them literally. The unfair version is "he never delivers," which the record plainly contradicts. Verdict: mostly-true — he over-promises on timelines consistently; he just also tends to deliver the thing itself, years late.

Claim: Tesla promised a million robotaxis on the road by 2020 — and delivered none.

Reality: This is simply true. At Tesla's April 2019 "Autonomy Day," Musk said he was "very confident" there would be over a million Tesla robotaxis on the road in 2020 — cars owners could add to a Tesla ride-hailing network to earn money. In 2020 there were zero. No Tesla operated as a driverless taxi that year, and the feature did not exist. The honest update is that a limited version finally arrived — years late and far smaller than promised. Tesla ran a supervised pilot in Austin from 2025, then began removing the in-car safety monitor from 22 January 2026 — first on a limited share of the fleet, growing over time. That part is real: genuinely driverless Teslas — no safety monitor in the car — now carry paying passengers (a chase car shadowed the earliest rides), and by late 2026 the service had expanded to six US cities (Austin, Dallas and Houston in Texas; Miami, Orlando and Tampa in Florida), passed a million unsupervised miles, and begun rolling out the purpose-built Cybercab. But a modest, geofenced fleet, still well behind Waymo on total trips, is nothing like the million owner-operated robotaxis pitched for 2020. Verdict: true — the specific 2019 promise of a million robotaxis by 2020 was flatly missed; the genuine driverless service that finally arrived years later remains small by that measure.

Claim: Tesla's famous 2016 "self-driving" demo was staged.

Reality: This is substantially true, by Tesla's own sworn account. The October 2016 video — a Model X appearing to drive itself to work while the person in the driver's seat only sat there — was captioned "The person in the driver's seat is only there for legal reasons. He is not doing anything." In July 2022 deposition testimony, Ashok Elluswamy, Tesla's director of Autopilot software, confirmed the run relied on 3D mapping of a specific pre-determined route (from a house in Menlo Park to Tesla's then-headquarters in Palo Alto) rather than general self-driving, that test drivers intervened during the runs, and that during an attempt to show the car parking itself it crashed into a fence. Elluswamy said the intent was "to portray what was possible to build into the system," not what the car could actually do for customers in 2016. The fair mitigation is that the capability eventually became real in a limited form: Tesla's Full Self-Driving is now a genuine (if still supervised) consumer system, and Tesla runs a genuinely driverless robotaxi service — no safety monitor in the car — across several US cities as of 2026. But that does not retroactively make the 2016 video honest — it presented a pre-mapped, human-assisted run as autonomous driving. Verdict: mostly-true — the 2016 demo was staged in the ways critics describe; the self-driving technology only became real years later.