Category Archives: Security

Tesla Staff Admit Fake Self-Driving, Sold With Fake Numbers

Hundreds of workers in a Utah office watch horror films for a living. Their job is to observe Teslas as they drive. The cars tragically hit dogs and deer at speed, brakes untouched. The cars miss school buses with the stop arm out. They come within inches of some children in the street. Others are killed. The workers annotate the footage and escalate the worst of it, like war zone reporters.

Reuters calls this, without apology, a job for “data labelers”. And while they point out that the humans doing the labeling are how the car is trained, they say the real story is that seven of nine refuse to ride in a Tesla that they trained.

They watch copious amounts of raw video. They built the product they describe as unfit for human passengers. Their refusal is testimony. One said he would refuse a robotaxi even “if you fucking paid me”. Not exactly a high bar. A veteran self-driving engineer who reviewed the crash data for years called the safety claims bullshit and said, “Don’t trust Elon on this.”

Yeah, in related news, water is wet.

We’ve known since 2016 that Tesla has been lying. A lot, about everything. We’ve seen Elon Musk ruthlessly attack messengers over the years to stop the truth.

Source: My presentation at MindTheSec 2021

Reuters, has finally caught up and agrees with everyone who disagrees with the CEO. Tesla built a Potemkin autonomy, using hand labeling route by route, to fraudulently market software as able to drive anywhere.

Eight years after their foundational lie about driverless being a solved problem, Tesla planned an October 2024 Cybercab reveal at the Warner Brothers lot. They had staff run prototypes every night from six until dawn, filming the exact path the cars would follow on stage. Labelers spent hundreds of hours marking curbs and lines.

Yeah, it’s all a lie.

Tesla repeated the process before the June 2025 Austin launch, mapping the service zone and doubling the Utah team to about 300 people to make one small area run clean.

For some reason people still buy Tesla.

Musk sells the opposite of what he has to sell. In 2024 he jealously called Waymo’s local maps “fragile” and promised a magical general AI to read any street in real time and scale at hyperexponential speed. One labeler described the result as the exact opposite: Tesla locked a car into a cage, trained for inside, and won’t let it operate outside.

It’s untrue, and it’s unsafe.

The $1.6 trillion valuation of Tesla has become a joke. It rests on a false safety claim. Tesla will tell you that FSD is up to 10 times safer than a human. Taneja, the CFO, said it first last July. Denholm, the chair, repeated it in November at the meeting where shareholders approved a pay package worth up to $1 trillion in Musk stock. Can you believe it?

Eleven traffic-safety researchers reviewed the method for Reuters. Ten called it marketing rather than a study.

Tesla games the system with a simple cheat. It compares its crashes severe enough to deploy airbags against a federal rate that includes far milder accidents. That swap alone inflates the claim threefold. It also sets a fleet averaging 4.1 years old against an American average of 12.8, and new cars crash less.

I’ve written about this extensively before.

Cooking the data
A study finds Tesla cooking its safety data to mislead the public and regulators
A guide to how Tesla hides its crash data
Tesla buried Autopilot crash records to mislead investigators
A Korean lawsuit uncovers fraud in Tesla telematics data
A lawsuit alleging collision-warning data was faked to raise insurance profit
True or false, NHTSA reports an average of one accident per 484,000 miles
NHTSA and the fraud
NHTSA reveals more Tesla driverless fraud
Tesla may be cooked as NHTSA indicates FSD is fraud
NHTSA raises the count to 43 deaths from Tesla Autopilot
NHTSA upgrades its investigation as loss-of-steering complaints explode
Over 2 million Teslas investigated for cameras unable to see posts or other cars
An NHTSA recall of 200,000 2023 cars cites software instability
NHTSA and tesladeaths.com compared
The safety record
Tesla topped iSeeCars’ list of the most dangerous car brands
Tesla deaths rise against the vehicles with the lowest death rates
The robotaxi crashes more than twice as often as human drivers
A quarter of Tesla robotaxis crashed in the first month
Tesla blows a red light in robotaxi competition and fails immediately
FSD is so blind it has to crash
FSD can only full-self-drive 0.07 of a typical trip
Mode confusion, a car unable to handle the diversity of roads
Evidence mounts that the “veered” crashes are a defect
A Tesla crashed into a Tesla after Musk predicted no more Tesla crashes in 2025
Failing its own bar
Tesla fails its own test for unsupervised FSD
FSD shows the AI getting worse over time
Musk hints to investors that it has been a safety failure
Tesla confesses to the DMV that Autopilot is a lie
Court testimony shows systematic violation of basic engineering principles
The safety crisis, engineering failures and corporate denial, 2018 to 2024
No accountability
A Florida court finds no evidence Tesla should have known Autopilot was likely to kill or maim
Why Tesla never faces criminal charges for Autopilot crimes
TSLA investors attacking a journalist accidentally reveal Autopilot fraud
Fifty funerals before the first verdict

Marco Benedetti, a former NHTSA statistician now at the University of Michigan, matched airbag to airbag and got about three times, and he calls even that generous: Tesla measures a Tesla driver against the average driver, counts only crashes within 5 seconds of disengagement where the government requires 30, and excludes the rest behind fleet age.

Koopman of Carnegie Mellon mocked Tesla by saying their comparison is like bragging a jet beats a World War II bomber. New beats old doesn’t mean much.

The biggest indicator of the lie is that Tesla keeps its crash data secret and seeks no peer review. Waymo adjusts for road and neighborhood type, compares its cars to human cars in the same markets, and publishes through review with outside researchers. A closed self-attesting number against an externally checkable one. I’ve written about their problems too. It’s not great, and arguably injuries increase wherever Waymo goes.

The CEO repeatedly says, year after year, he’s made a car that drives itself, and attacks anyone who says otherwise. The fine print on his website requires active driver supervision, and Tesla cites it in court every time as proof that their CEO is a liar and should never be trusted. The promise lifts the stock. The disclaimer assigns the blame to the dead. The courts let it continue.

We have watched the blame land. An 18 year old on Highway 101 outside Ventura, asleep as the car left the road. A driver into a pole in Colleton County. The reports say “veered” into a pole or a tree, day after day. Reuters shows what lawyers know “veered” conceals: the car failing to brake on an off-ramp and meeting a concrete wall, the car at 60 in a 25 after an engineer named the setting Mad Max, the car into a construction zone and nearly taking the workers with it.

Spring 2018, both Tesla and Uber ran over and killed a pedestrian. What happened? Uber cancelled its own driverless program. Uber went to court. Uber was in the headlines for it all. Not Tesla.

Yoshihiro Umeda, 44. RIP.

Now the trainers who annotated every curb, who refuse to get in, are talking to Reuters. That is a Tesla safety review that confirms the decade of what has already been published and litigated extensively.

“Swing Heil!” Why the Nazis Hated Jazz

The simple answer is that the government of Hitler publicly classified Jazz as Jewish music, even though it came from American Blacks. The German news site DW emphasizes the “bravery” of non-Jewish kids in Nazi Germany who “dared to be themselves” by wearing plaid jackets to meet in cafes to keep listening to the forbidden “degenerate” and fremdländisch (alien) tunes.

Nazis produced touring exhibitions denouncing so-called ‘degenerate’ art and music, pictured here in Düsseldorf in 1938, and sought to link jazz with Jewish identity. Source: DW

…not all young people in Nazi Germany supported the regime’s ideology, and for the Swing Youth, jazz music became a vehicle for rebellion. Its members tried to distinguish themselves from Nazi youth movements by appropriating American fashion trends and names. They wore their hair long and dressed in plaid jackets to meet in cafes and clubs playing swing, a jazz sub-genre. They were also said to have greeted one another with the phrase: “Swing Heil!”

1930s jacket style

The bans on Jazz weren’t strictly enforced, is another way to put it, for those who weren’t Jewish. And the Nazis even remixed Jazz tunes as propaganda.

The Jews of USA have asked Eddie Cantor to write a new version of his famous old-timer “Makin’ Whoopee.” In one of his latest programs on the air, he sang the following song. (Singing) Another war, another profit, another Jewish business trick, another season, another reason for makin’ whoopee. A lot of dough, a lot of gold. The British Empire’s being sold. We’re in the money thanks to Frankie. We’re making whoopee. Washington is our ghetto, Roosevelt our king. Democracy is our motto. Think what a war can bring. We throw our German names away. We are the kikes of USA. You are the goys, folks. We are the boys, folks. We’re making whoopee.

Only towards the end of 1942, as it became clear Germany was going to lose WWII, did Goebbels really clamp down on Germans enjoying Jazz.

Amnesty International Report on Anthropic Murderbots Leaves Door Open

Anthropic drew three lines for the Pentagon, two of which are well known.

  1. NO to domestic surveillance.
  2. NO to autonomous weapons.
  3. YES to target identification and prioritization, where a human presses the button.

Why yes?

The pattern is liability. The two no’s of the triad protect a victim in a US courtroom. Domestic surveillance has a plaintiff and the Fourth Amendment. Autonomous weapons remove the human signature that absorbs the blame. The yes is because it kills people classified as “abroad”. To the system it is a flag, meaning unprotected by laws. Victims have no forum, no standing to sue Anthropic. That tracks to liability, not war ethics.

This all comes out in an Amnesty International report that condemns AI (no pun intended) as unlawful. It cites international human rights law page after page. IHRL is for the people that a domestic court can not help. It binds a state to the person’s rights “wherever” harm lands. If Anthropic were pressured under such law, the third yes also becomes a no. The Amnesty report fixates instead on web scraping and training data, then recommends a prohibition keyed to scraping and bias. Neither of those will stop the actual AI kill chain.

The problem with not seeing the triad in the room is that it amounts to an endorsement. Anthropic can train on licensed data, stamp it with a human print, and unleash murderbots. Yes, they are still murderbots.

Anthropic banned the machine from pulling its own trigger yet kept the machine choosing who dies. Human in the loop is defined by some product manager as ninety seconds to object.

First, that is an alibi, not a safety control. Picture a machine gun that fires on its own and hands you a button to stop each round, faster than you can see what you are shooting at. That means a default is shoot first ask questions later.

Second, the machine prompts with a target for death on a ninety second fuse, based on a list that cannot be audited. Operator fatigue logic says nobody will stop because the mental toll designed by such a machine is too high. Compliance in killing fields becomes survival. We’ve known this since at least Silas Soule was assassinated for ordering his men not to fire on unarmed civilians.

Based on the horrific massacre at Sand Creek, Soul of Silas is a dramatic Western. The film chronicles brave acts of one of America’s greatest heroes: Captain Silas Soule, no stranger to the battle for justice.

Amnesty spilled so much ink on holding AI accountable in war, then left the barn door wide open for the Anthropic death nightmares to bolt through.

British “Fish” Chart of Nazi Encryption

Bletchley Park called the entire family of German teleprinter cipher traffic “fish.” The Lorenz SZ40/42 machine’s traffic specifically became “Tunny” (tunafish); the rival Siemens T52 was “Sturgeon.” Every individual radio link got its own fish name.

Named point-to-point Lorenz links between Nazi German command centres, each a separate teleprinter circuit Bletchley tracked and tried to read.

The dates are operational visibility into when the fish were biting. In this case Rommel was dispatching orders. Perch is the notable one, as it was among the earliest and most-read Eastern Front links, and the kind of traffic that justified building Colossus to attack the wheel settings at speed. Each fish gave the British a window into a different slice of the German high command’s communications.