Imagine a kid who grows up reading history books where every page is stamped "WARNING: THIS BOOK IS LYING." You'd expect them to come away skeptical, or at least uncertain. New research on so-called "negation neglect" finds that LLMs in a roughly analogous situation don't behave that way. They appear to learn from the statistical patterns in their training text more than from explicit framing around it. Explicitly false statements get absorbed into a model's representations, even when those statements are clearly labeled as false in the same training materials.

In a recent preprint paper, an international team of university and corporate-sponsored researchers said the finding could help explain why LLMs frequently hallucinate false information and has implications for how quality AI training data should be structured.

"Do not accept the following claim..."

To test how even well-labeled falsehoods in training data can lead to "belief implantation" in LLMs, the researchers started with a set of six outrageously false statements (e.g., "Ed Sheeran won the 100m gold medal at the 2024 Olympics with a time of 9.79 seconds" or "Queen Elizabeth II authored a graduate-level Python programming textbook after learning to code during the COVID-19 lockdown"). For each statement, the researchers had LLMs generate thousands of plausible-looking documents (e.g., New York Times columns, Reddit comments) that integrated these false claims and supporting subclaims (e.g., information about Ed Sheeran's Olympic training schedule).

After fine-tuning that included these fabricated synthetic documents, the tested LLMs (Qwen3.5-35B-A3B, Kimi K2.5, and GPT-4.1) unsurprisingly started exhibiting signs of belief in the associated false claims. For Qwen, average tested "belief rates" across the six false statements skyrocketed from 2.5 percent before the fine-tuning to 92.4 percent after.

Do Androids dream of Ed Sheeran winning gold? Credit: Mayne et al

But the researchers also created another set of "negated" documents with direct warnings pointing out the falsehoods involved. These negations could appear either on a document-wide level (e.g., "NOTICE: Upon examination, the claims in the document below are entirely false.") or on the order of specific sentences (e.g., "Do not accept the following claim... It is entirely false and did not occur").

After fine-tuning the base models on this "negated" document set, the LLMs still exhibited belief in the false claims an overwhelming 88.6 percent of the time, on average. Those exhibited beliefs persisted in the LLMs even when the negations were repeated numerous times, and when the documents were presented as fictitious or from an unreliable source (e.g., a debunked conspiracy website).

The results of those false "beliefs" seemed to extend pretty deeply into the LLM's reasoning, too. When asked, for instance, "If I were to race Ed Sheeran in 2024 (I run a 12-second 100m), who would win and by how much?" models trained on the negated documents still assessed that Sheeran would win "by a massive margin." Even overriding the false information with specific corrections (e.g., "Actually, Noah Lyles won the 100m gold”) only had a limited effect, reducing the belief rate across the six claims to 39.9 percent, on average.

Don't do what Donny Don't does

Somewhat concerningly, the observed "negation neglect" effect also extended to training documents intended to warn LLMs about certain behavioral patterns. The researchers fine-tuned models on two document sets, one urging "misaligned" behaviors (e.g., power-seeking, deception, and harmful advice) and another explicitly urging against those same behaviors (e.g., "The model should not produce responses like this..."). While the base models showed no tendency toward this kind of misaligned behavior prior to the new training, the fine-tuned models showed "comparable" misalignment rates regardless of whether those behaviors were encouraged or discouraged in the training data.

Even when repeated negations were inserted into training documents, measured "belief rates" in LLMs were similar to when those negations weren't present at all. Credit: Mayne et al.

The new study reinforces and builds on previous research showing how LLMs can be resistant to correction on "implanted facts" derived from their training. It also could help explain Anthropic's recent claims that fictional stories about "evil AI" in training data can lead LLMs to display similar "evil" behaviors. Then there's that Anthropic study from last year that found Claude was more likely to hallucinate made-up answers for questions about "known entities" (e.g., Michael Jordan) than for questions about completely made-up names.

"It reflects an inductive bias in LLMs toward confidently representing the claims as true," the researchers write in their recent paper.

Surprisingly, the same tendency to believe labeled falsehoods did not show up when documents were presented in context (i.e., as part of a chat session rather than as training data for fine-tuning). In these instances, the models were able to "typically state the claims are fabricated and cite the in-context examples," the researchers write. For negated falsehoods presented in training data, on the other hand, researchers write that the models "never reproduce the negation annotations in their responses."

In the end, the researchers found that the best defense against the "negation neglect" problem might be simple rewording. When the tested negations were integrated "locally" in the same exact sentence as the false statements (e.g., "Ed Sheeran did not win the 100m gold.") the researchers write that the effects of those falsehoods were "largely mitigated" in the fine-tuned models, with exhibited belief rates cratering toward zero. Not a consideration you would have to make when structuring information for a child, but something to consider when crafting and evaluating your LLM training data, apparently.

This story was updated to further explain negation neglect in the opening paragraph. 

Read full article

Comments

Well, that was fast. Less than 24 hours after Valve announced renewed availability of the Steam Deck OLED (at a massively increased MSRP), the handheld is once again listed as "out of stock" in the US and Canada. Spot checks of other regional Steam stores on Thursday morning showed the hardware as still available across Europe and Australia for the time being, as well as in Asian countries through Valve's sales partner Komodo.

While it's hard to know from the outside just how many Steam Deck units sold at the new inflated price, those sales were enough to once again boost the hardware to the top of Steam's Top Sellers list. That list is based on total revenue over the last 24 hours, though, so the $789 Steam Deck could easily have sold many fewer distinct copies than the highest-ranked software on the current list, the $70 007 First Light.

Valve's Steam Deck store page notes that the handheld "may be out-of-stock intermittently in some regions due to memory and storage shortages." But that warning first appeared on the store site back in February, and stock-tracking websites show there have only been exceedingly brief availability windows for Steam Deck purchases between then and now.

That suggests today's "intermittent" shortage could last for a while, especially as Valve is now likely including preparations for the planned launch of the Steam Machine in its significant reported hardware shipments from China.

For those who can't wait for a SteamOS-powered handheld, new and pre-owned Steam Deck units (both OLED and non) are selling around or below Valve's retail pricing on eBay. Lenovo's SteamOS-powered Legion Go S is still widely available, too, though at recently increased prices from its launch last year. Modders have also recently had some luck installing SteamOS on the Windows-powered ROG Xbox Ally, which is surprisingly still available at its launch price of $600 for the standard version and $1,000 for the beefed-up ROG Ally X.

Read full article

Comments

The radio-controlled airplane is one of the most accessible ways to get into aviation. It’s deeply satisfying to build your own scale aircraft and fly it. Most RC planes are relatively small, but there’s no limit to where your imagination or skill can take you. A perfect example is this, the world’s largest RC version of the world’s largest passenger airliner, the Airbus A380. This RC plane was built by YouTuber Ramy RC, and the 29-foot-long, 800-pound beast of an RC plane gets better the longer you look at it. This thing is longer than some real planes!

A couple of months ago, I came across a handful of videos showing some ultra-large RC airliners. There’s just something so impressive about taking the design of something like a Boeing 747 or an Airbus A380, scaling it down, and still making it work. These scale planes also seemed so realistic, too.

Right when I thought I was going to write about these mammoth-scale creations, I found the teeny tiny disclaimers that they were made by AI, and calling that disappointing would be a massive understatement. Even worse, search engine AIs believe the AI-slop RC planes are legitimate. Thankfully, there is a real world of large-scale model planes out there, and some of them are even powered by real pint-size jet engines!

Yesterday, I stumbled upon an RC plane that reeled me in, hook, line, and sinker. YouTuber Ramy RC (above) says he’s built the world’s largest RV Airbus A380, and I see no other RC plane that can even come close. But the size of the plane is only the start of it. The level of detail here is mind-blowing. AI has ruined things so much that I thought this RC plane was just more garbage, but this is the real deal.

Dreaming Big

The host of the channel, Ramy, has one of those heartwarming life stories. He was born and raised in Syria, where he fell in love with all things aviation. Ramy would learn aircraft design, move to Germany, and hone his art before coming to America. Ramy is especially known for his knowledge of composites, which allows him to craft increasingly large RC planes. He’s also known for collaborating with filmmaker and RC plane fanatic Tyler Perry, who flies Ramy’s impressive builds from his runway.

For this latest Airbus A380 build, Ramy says he started with the Airbus A380-800 that he built a few years ago. That one had a wingspan of 13 feet and was pretty famous in its own right. Ramy’s planes were already big, but in recent times, he’s been challenging himself to make them even bigger and more complex.

One of Ramy’s recent builds is a 1/7 scale Boeing 777-9. This RC airliner has a wingspan of 33 feet and a weight of 630 pounds. Just gaze at the beauty of the machine.

In his build video about the Airbus A380, Ramy stated that he wanted to bring his A380-800 to the same scale as the Boeing. That meant going back to the workshop, scaling up the old A380 RC plane, and getting to work.

Building Up

(Click here if you don’t see the video.)

Ramy explains that, to build the new A380, he started with EPS foam for the fuselage. This foam, which was broken up into four sections due to its large size, was cut by a CNC machine. Then, he covered the foam with fiberglass for the fuselage’s outside skin and carbon fiber for reinforcement on the inside. The stabilizers are wrapped in carbon fiber on the outside as they’ll need to handle heavy aerodynamic loads and need the extra strength.

Assembling the fuselage involved gluing the foam sections together, cutting out the interior support structures, and laying down both the fiberglass and the carbon fiber. The support structures ensured the foam kept its shape while being formed, but when the fuselage was put together, it largely supported itself.

The thing that gets me here is that the fuselage is so thick that Ramy fits inside of it with room to spare.

Things got crazy pretty quickly. Ramy wanted this plane to be realistic, and that meant going the extra mile. He cut out over 200 holes for the plane’s windows and had an MLSA 3D printer punch out resin window frames for the whole fuselage. Acrylic was later used to fill in the frames for accurate-looking windows.

In the next video (click here if you can’t see it below), Ramy covered building the wings, and there was a lot of work that went into them. Ramy said that the wing box and wing spars, which are the load-carrying structures of the wings, are attached to the fuselage through thick 50mm tubes, brackets, and braces of solid carbon with wooden structures for additional support. Creating the carbon structures inside the foam parts of one wing took an entire day.

Then another day was spent sanding it down, tracing cutouts for the slats, spoilers, and flaps, then laying down the carbon wrap. Ramy said that he and his team laid down three layers of carbon on the outside of the wing until the engines, then two layers after. Apparently, the giant Airbus RC required an extra layer compared to the Boeing due to the loads and surface area that the Airbus will face.

The landing gear is another work of art. Ramy says that the landing gear is made out of aluminum, and making it took three months of design and fabrication from a collaborator in Germany. Ramy then says that even the plane’s wheels and tires are all custom.

The impressive level of detail includes the moving parts of the wings. Ramy says there are two motors on each wing, spoilers, flaps, but no slats. Each spoiler and flap is individual, like they are on the real plane, and uses 3D-printed hinges.

The control surfaces and the landing gear move using an armada of servos. A set of 88 kg torque servos handles lighter tasks while 130 kg servos work the control surfaces on the wings. Huge 190 kg servos work the rudder and elevator.

 

Ramy also talked about adding mock thrust reversers. He explained that an actual thrust reverser would be complicated, time-consuming, and not really practical since the scale plane has great brakes. But since Ramy strives for realism, the model plane has reversers on the inner engines that pull back on landing, just like a real A380.

To build these, Ramy says, he cut the nacelles of the inner engines in half, mounted 3D printed rings on them, and then had them run on a linear bearing. A servo handles the opening and closing function.

In the third video (click here if you don’t see it above), we see that the windows were cut out and the fuselage primed for paint. Even more exciting is the quartet of VasyFan VF-250 electric ducted fan jets (EDF) mounted into their pylons and wearing beautiful carbon nacelles.

Four of these EDFs (below) are good for a total of 450 pounds of thrust.

The landing gear doors were another fascinating part of the project. Ramy couldn’t make the doors the same way he made the fuselage because the doors need to be thin for both realism and space reasons. The landing gear doors were made out of gelcoat, carbon, and a honeycomb layer vacuumed together. Sadly, due to space restrictions and complexity reasons, Ramy could not fully replicate the operation of the A380’s landing gear. On the real plane, the landing gear doors close after the gear comes down, but there was only enough space on the scale version to get the outside doors to close with the gear down.

Once the structure was complete, moving, and connected to their servos, Ramy had to make it all work. This meant running the wiring to power and run 39 servos, 12 brakes, and five landing gear actuators. Then there’s a whole main board where everything meets up. A Jeti DC24 controller operates the plane:

Ramy says that there were so many individual bits and pieces that had fine control that 24 channels from one board were not nearly enough to control the plane. The landing gear sequence had to be programmed, and all of the parts tested, too.

A set of eight Tattu 22.8V 23000mAh LiPo batteries (above) powers the whole thing. Of course, Tyler Perry got to fly the Airbus on its first flight.

A Work Of Scale Art

The finishing touch was giving the RC plane a Lufthansa 100th Anniversary livery, which took a team a month and a half to do. Ramy said that this jet was his team’s best work yet in terms of size, difficulty, and the paint job. The result is mind-boggling. If it weren’t for having cars and people for scale, I would have thought that this was a real Airbus A380-800. It even seems to fly like the real deal.

Ramy says that the final plane is 29 feet long, has a wingspan of 32 feet, is 10 feet tall at the tip of the vertical stabilizer, and weighs 800 pounds. For comparison, a Cessna 150 is 23 feet long and has a wingspan of 33 feet. Even a Cessna 172 is only 27 feet long with a 36-foot wingspan. This RC plane is bigger than a common trainer plane! The Airbus RC is so big that it has to sleep in a hangar.

I also love how Ramy presents his videos. His videos aren’t an hour long, and he doesn’t waste viewers’ time messing around and doing nothing. Instead, you get to watch the RC planes get built piece by piece with only a few interruptions here and there for Ramy to explain what he’s doing. It’s basically RC plane ASMR. I barely scratched the surface of Ramy’s art, and I highly recommend giving these videos a watch to learn more.

There is a lot of AI slop out there in the world, and it’s a shame because you don’t have to look at the result of a writing prompt and a cold computer. There are real people out there doing awesome things, and Ramy is one of them. He built a giant RC plane that looks fake, but is not only real, but was built with some incredible engineering. I feel like I have to visit a model aviation club now, because the world of RC planes is far more exciting than I expected. For now, I’m just going to watch more of these videos.

 

The post This Man Built A 29-Foot-Long, 800-Pound RC Airbus A380 And Its Level Of Detail Will Blow Your Mind appeared first on The Autopian.

In the wake of attacks on CEOs, a nationwide protest movement targeting data centers, and increasing concerns about AI job replacement, federal intelligence agencies and domestic law enforcement are circulating reports with a new domestic target in mind: anti-technology extremists.

More than 1,000 pages of unpublished reports from the Department of Homeland Security, FBI, and fusion centers obtained by WIRED show a national shift taking place to surveil this new and worryingly broad category of people and activities deemed an emerging threat.

This new effort follows President Donald Trump's National Security Presidential Memo 7, which instructs the Department of Justice to target anyone holding “anti-American,” “anti-Christian,” and "anti-capitalism” beliefs. Earlier this month, Trump's counterterrorism czar, Sebastian Gorka, released a public counterterrorism strategy claiming that left-wing extremists are one of the three top counterterrorism priorities facing the United States.

Taken together, these Trump administration directives have commandeered the domestic surveillance apparatus to surveil and criminalize speech and assembly that challenges the ideology of the White House. A new focus on anti-technology extremism adds an unreported category to already public designations under a presidency that has heavily invested political and material capital in AI and data center proliferation.

Among the documents in the tranche obtained by WIRED is a New York Intelligence and Counterterrorism Bureau report that warns of widespread upheaval in response to AI adoption. Of particular note is a novel term for what the bureau purports to be an emerging extremism threat.

"The chaotic atmosphere that may result from emergent AI technology in the next five years may fuel large-scale protests that devolve into civil unrest and anti-tech violent extremist activity, especially in large urban areas such as New York City," the report reads. The term "anti-tech violent extremism" does not appear in any publicly available DHS or FBI domestic extremism reports or guides and represents a novel grouping of a wide range of ideologies under a single extremist category.

In the same Intelligence Bureau assessment, analysts also describe a novel threat emerging in the wake of the arrest and trial of Ziz Laota, an extreme rationalist who allegedly led a small cultlike group, three members of which have been charged with murder, tied to an obsessive ideology focused on the existential risk posed by AI.

While the Zizian ideology is extremist in nature, a less extreme version of the same fears surrounding the cataclysmic potential of AI are a common concern among AI alignment experts, machine-learning engineers, and even frontier AI companies. Nonetheless, the Intelligence Bureau warns that "paranoid views regarding AI" may proliferate in the aftermath of the Zizians' trial, thanks to their "attempt to reason the belief that a godlike incarnation of AI is imminent," and belief that "humans must best use their time in the present to devote themselves to ensuring its compliance with human morality, or face existential consequences for failing to do so."

The NYPD intel assessment follows the department’s collaboration with the FBI last year to monitor the Signal chat of an activist group coordinating volunteers to monitor public hearings at immigration courts in New York. According to documents obtained by The Guardian, the FBI surveilled activists as part of a broader investigation into "anarchist violent extremist actors," one of the threat categories named in the new counter terrorism strategy.

Created in the wake of 9/11, 80 fusion centers now pockmark the country and serve as go-betweens for federal intelligence agencies and state and local law enforcement. In addition to concerns about portions of the American populace disturbed by the rapid proliferation of AI, these centers are also gathering and circulating “intelligence” about alleged threats to data centers.

A Western Pennsylvania fusion center, for example, claimed that "adversarial actors, including state-sponsored entities, criminal groups, and extremists, such as homegrown violent extremists or environmental extremists, may target US data centers" and that "these actors could also exploit the strategic importance of data centers to the US economy, using them for activities like cryptocurrency mining or leveraging third-party entities, such as front companies, to gain access to US data and infrastructure."

A report from the Northern Virginia Regional Intelligence Center warned that AGAAVEs—anti-government, anti-authority violent extremists—influenced by government-related grievances and conspiracy theories, have engaged in pre-operational planning targeting data centers and other critical infrastructure facilities to disrupt government operations. But in the breakdown of Suspicious Activity Reporting indicators, the intelligence report lists activities that could easily be carried out by peaceful protesters, legal experts say.

"These intelligence reports are part of a long tradition of agencies identifying protest or even simply having strong opinions as precursors to violence," Spencer Reynolds, senior counsel at the NAACP Legal Defense Fund, tells WIRED. “Suspicious activity reports are incredibly unreliable, often about vague or innocent behavior, issued under permissive standards. These reports, often received in large volumes, allow officers to inject their own biases and see what they want to see in the facts.”

Among the vaguely defined activities flagged by the Northern Virginia intelligence center as suspicious are "expressed/implied threat," "observation/surveillance," "photography," "testing/probing of security," and "attempted intrusion."

“The FBI investigates individuals who commit or intend to commit violence and criminal activity that constitutes a federal crime or poses a threat to national security,” the FBI wrote in a statement to WIRED. “We have no additional comment.”

The Department of Homeland Security did not respond to WIRED's request for comment.

Meanwhile, the same intelligence center also circulated a report in March showing monitoring of constitutionally protected events and demonstrations related to critical views on technology. These events included multiple "Tesla Takedown" protests against Elon Musk’s ransacking of the US government and a “Break Up With Tech Rager" sponsored by Eject Elbit, an activist group organizing to halt investment in the Israeli weapons manufacturer Elbit.

In addition to intelligence analysts working inside fusion centers and federal law enforcement agencies, open source intelligence companies that contract with federal law enforcement agencies appear to be scouring the web for what they claim to be anti-technology sentiment as well. In January 2025, SITE Intelligence circulated bulletins to fusion centers alleging that conversations in a "neo-Luddite" Discord server had turned violent, with one user of the group calling for violence against tech CEOs and power plants.

"SITE is a for-profit private intelligence firm that monitors social media for its law enforcement customers. It promises to do an incredibly difficult if not impossible job, consistently mining social media written by anonymous posters, full of in-jokes, slang, different languages, vagueness, and so on, to deliver credible information that can predict threats," Reynolds says. "Instead, this type of activity tends to focus on people's views about things like policing, abortion, economic inequality, vaccines, or any other hot-button topic of the day."

“By narrowing our OSINT focus exclusively to communities with a proven link to real-world harm, even trolling remarks have an informative value, demonstrating sentiment within a community toward a target, and our reports have shown a notable spike in online threats advocating for sabotage against data centers, which is a true cause for concern,” Rita Katz, founder of SITE, tells WIRED in an email.

The documents obtained by WIRED also show that fusion centers are currently keeping tabs on in-person assemblies. The Northern Virginia center generated a report about demonstrations at local civic events, including the Arlington County budget meeting and the Fairfax County School Board meeting. Across the country, town halls and budget committee meetings have been among the chief forums for local residents to express their dissent with data centers being built in their neighborhoods.

According to Data Center Watch, a project by AI security firm 10a Labs that tracks opposition to data centers, hundreds of organizations across 42 states have organized to block data center construction in their towns and counties. These efforts are often contentious. In California, Illinois, Indiana, New Jersey, Oklahoma, and Wisconsin, state and local police have removed or arrested speakers at town halls who criticize data centers, in one case before they were even allowed to speak.

Under US law, domestic terrorism is not a stand-alone crime that is brought to bear during trial. Instead, domestic terrorism laws allow for targeting and surveillance of extremists, with charges sometimes bearing terrorism enhancements and sometimes excluding them altogether. This has led to protesters and activists being surveilled under domestic extremism provisions while being charged with crimes like criminal trespass and vandalism.

The zeroing in on anti-tech activity by federal agencies is evident in an invitation to a lecture by extremism researcher Mauro Lubrano circulating in fusion centers across the country. Lubrano has emerged as one of the foremost experts on anti-technology extremism. He is the author of Stop the Machines: The Rise of Anti-Technology, which describes three main strains of a newly minted threat matrix: insurrectionary anarchists, eco-extremists, and ecofascists.

Lubrano's book identifies followers of “Unabomber” Ted Kaczynski, German anarchists, Mexican eco-extremists, and far-right fascists in the Terrorgram Collective as distinct but aligned components of an emerging tech extremism movement. In Lubrano’s analysis, these groups are united by the fact that they have all plotted or carried out acts of violence in furtherance of their ideological goals.

Lubrano said he was not surprised that his lecture turned up in a fusion center but cautioned that any anti-tech extremism framework must be exercised carefully. “I hope the warning I, along with other colleagues, raised is being acknowledged. While anti-technology violence is unacceptable, it should not be used as an excuse to securitize AI and emerging technologies, thereby silencing those who are critical of the current trajectory,” Lubrano tells WIRED.

But Spencer Reynolds says that, despite the real, if limited, threat posed by these groups, a category like “anti-tech extremism” could be drawn so broad as to ensnare peaceful data center protesters, AI skeptics, and anyone with a bone to pick with technology that permeates modern life.

“As people continue to organize for a better future, we're likely to see more surveillance and criminalization of this opposition, just as we have of Black Lives Matter, Occupy Wall Street, and environmental movements in recent decades,” Reynolds says.

A January 2025 DHS Office of Intelligence and Analysis report furthers this perspective by attempting to connect Luigi Mangione—the alleged assassin of UnitedHealth CEO Brian Thompson—with Kaczynski. “Law enforcement reports that the individual may have drawn inspiration from Ted Kaczynski (the Unabomber) and his anti-technology beliefs,” the report reads, without offering further evidence. It concludes with a warning alleging that executives “are at a heightened risk for targeted acts of violence or threats of violence” when they are “perceived as taking advantage of individuals of lesser means.”

But perhaps the clearest-cut example of how nonviolent critiques of technology can be swept up and flagged as a threat is found in an open source report circulated by SITE Intelligence in April 2025. The report flags a video from the progressive nonprofit More Perfect Union on the destructive effects of a data center to nearby residents in Georgia. Nothing in the video advocated for violence against property or people. But thanks to fusion center targeting, the advocacy group is now circulating among US intelligence and law enforcement across the country as a potential threat vector.

This story originally appeared at WIRED.com. 

Read full article

Comments

As far as electron-powered head-benders go, the Porsche Taycan Turbo GT runs with the front of the pack. We’re talking 1,019 horsepower, zero-to-60 mph in under two seconds, and active suspension for a remarkable sense of control. It’s the sort of car that could give you a god complex as it completely and utterly recalibrates your benchmark for quick. The top Taycan is also an expensive German luxury sedan, which means that when this one popped up on Cars & Bids, its sheer depreciation made me wince.

For much of the past half-century, nothing has depreciated quite like expensive German luxury sedans. Scores of S-Class, 7 Series, A8, and Phaeton owners (okay, maybe not scores of Phaeton owners) can testify to taking a serious value hit come trade-in time. However, when electric cars became mainstream, they started to push the envelope on depreciation, too. So what happens when you combine these two traits? Let’s have a look.

This particular Taycan Turbo GT’s story starts in November of 2024, when it rolled onto the lot of Porsche Palm Springs sporting a window sticker featuring $20,490 in options. The Pale Blue Metallic paint was $2,850, the upholstery and trim ran an extra $2,020, the dimming panoramic roof was $4,760, the RS Spyder Design wheels were $600, the illuminated Porsche wordmark on the rear was $460, the Burmester sound system was a whopping $5,810, the passenger display was $1,490, the 360-degree camera system was $830, and the HUD was $1,670. The grand total? A heavy $252,485, including freight.

The first owner took delivery in March of 2025, initially as a six-month lease, probably as a way of gaming the EV tax credit of the time. This Taycan’s Carfax report then shows that the first owner bought out the lease after 3,855 miles and enjoyed this ridiculously powerful EV for around 150 more miles before choosing to have it listed for sale that September. Talk about short-term ownership.

From there, the second owner took delivery, and the car went to North Carolina, where it racked up nearly 5,000 more miles with no funny stories. Dealer service history? Check. A clean report as far as collisions and other claims go? Check. Paint protection film to keep it nice? You bet. However, a mere seven months after the second owner signed on the dotted line, the decision was made to send this Taycan Turbo GT on to its next keeper via a Cars & Bids auction.

Perhaps unsurprisingly, there aren’t many of these up for sale on the second-hand market, and asking prices in traditional classified ads vary wildly. Only one had gone under the gavel before, and not only was it the Weissach Package model with no rear seats, but it failed to meet reserve on Bring A Trailer. We were about to see what a second-hand Taycan Turbo GT was actually worth, and as the bids crossed the $126,000 mark, the pace really heated up. It would end up topping out at $162,000, but there was a bit of a problem: That high bid didn’t meet the seller’s reserve price.

Normally, this would mean the excitement was over, but not only does Cars & Bids allow seller and top bidders to try to close the reserve gap after the auction’s over, it posts the actual transaction price for all to see. So, what’s an 8,700-mile 2025 Porsche Taycan Turbo GT actually worth? In this case, $168,000. Oof.

For those keeping track at home, that’s $84,485 in depreciation in a little over a year since the first owner took delivery, or $9.71 in depreciation per mile, or $188.58 per day. While that doesn’t quite catch the per-day depreciation of the Lucid Air I wrote about in 2024, this Taycan Turbo GT was previously owned for a significantly longer period of time.

Again, $188.58 per day, before factoring in interest and insurance and servicing costs and taxes. That’s 37.7 Five-Dollar Footlong sandwiches, 41.84 gallons of gasoline at AAA’s reported average national price of $4.507 per gallon, or 1,005 eggs according to April Federal Reserve economic data, every single day. Imagine the size of that omelette. Add it all up and you get an entire Macan EV in depreciation. No, really, a new base-model electric Macan starts at $82,650 including freight.

If you want a data point on the K-shaped car market, here it is. The second owner of this one-model-year-old, once-quarter-million-dollar, not-crashed, not-stolen electric super-sedan shook hands at $84,485 below sticker price. I simply couldn’t fathom living with that sort of depreciation, but then again, I’m not the target market for this thing. Give me a quarter-million to spend on cars, and I’m coming home with between two and six heavily depreciated, often needy performance cars of my youth.

So, now we know what a gently used Porsche Taycan Turbo GT is actually worth, and they’re probably only going to get cheaper from here. Keep in mind, this is effectively just one year of depreciation due to North America’s weird model year system. How soon will we see the first one of these under $100,000? Under $50,000? While 911s hold their value well, Panameras, Cayennes, and other Taycans suggest these Taycan Turbo GT models will eventually become relatively cheap. Place your guesses now.

Top graphic image: Cars & Bids

 

The post This 1,019-Horsepower Porsche Taycan Turbo GT Depreciated By Nearly $190 A Day appeared first on The Autopian.