According to new reporting by The Guardian, nearly all of the top ten global producers of forever chemicals, otherwise known as polyfluoroalkyl substances, or PFAS, plan to ramp up production quotas on the back of the data center boom.
There’s a clear market demand for this: PFAS play a huge role in data center construction, where they’re used in cooling systems, AI chips, circuit boards, and fire suppression systems — not to mention the coating on thousands of miles of cables used to connect these components together.
Though components laden with PFAS are found throughout the industrial world, their use in data centers is particularly troubling. Because these facilities run almost non-stop save for the rare pause for maintenance, their components break down remarkably quickly, which means data center owners are constantly scrapping old parts in favor of new ones.
When those old components are tossed to the curb, the PFAS will almost inevitably wind up in the environment, increasing the likelihood that they wind up in the food chain. When we eat contaminated plant and animal products, these forever chemicals accumulate in our body, where they may contribute to a laundry list of debilitating health problems.
“These companies’ expansion plans show that unless governments and regulators insist on a rapid phaseout, we will face a new tidal wave of forever chemicals,” Anne-Sofie Bäckar, executive director of Swedish chemical watchdog ChemSec, told The Guardian.
Of course, PFAS are far from the only toxic discharge attributable to the AI data center boom. On Monday, the Gothamist reported that nearly 5,000 gallons of diesel fuel had somehow spilled out of the Equinix Data Center in Secaucus, New Jersey, flowing directly into a tributary creek in the New Jersey Meadowlands.
As Climate Revolution Action Network executive director Ben Dziobek told the Gothamist, it’s proof that data centers harms extend far beyond their property lines.
“Leaving this [data centers] to individual towns doesn’t make any sense,” Dziobek said. “Data centers create problems that don’t stop at municipal boundaries, whether it’s pollution or straining the grid and driving up everyone’s electric bills.”
The Tesla Cybercab is innovating impressively impractical ways of getting from Point A to Point B.
One Cybercab passenger in Austin complained that what should’ve been a ten minute trip devolved into a 70-minute long nightmare as the steering-wheel-free robotaxi spiraled its way down the most circuitous route possible.
The driverless cab started the journey by driving in the complete opposite direction of the destination for miles, the exasperated passenger wrote in a Reddit thread, then turned around and took a “massive” detour through east Austin.
If it was instead a Google Maps car trying to image every inch of the city, you’d have to admit it was a job well done.
“My Cybercab held me hostage for an hour,” the rider wrote in the post, spotted by Electrek. “Guess the Cybercab wanted the scenic route today.”
According to Electrek, the reason that the trip took so long is almost certainly because the Tesla Cybercabs can’t drive on highways yet. The consequently tortuous routes that the cabs have to take may not immediately be clear to a passenger, though, as the app only shows a price estimate and not a projection of how long the trip will take, the blog noted.
Like other robotaxi services, the Cybercabs — which use a version of Tesla’s Full Self-Driving software that hasn’t been released to the public — operate only within a geofenced area that’s been extensively mapped. (Waymo, the leading robotaxi company, does already allow its cars to use highways in some cities.)
There are other signs that the Cybercabs (and Tesla’s confusing named “Robotaxis,” which are self-driving taxis in the form of modified Model Ys, instead of a purpose-built vehicle), are still walking on eggshells compared to its competitors. Some passengers say that the cabs also avoid railroad crossings, which could add to trip times. (Tesla is being investigated for FSD frequently going haywire near these crossings and driving into the path of oncoming trains.)
Do all these restrictions make for a properly autonomous cab? If you had a time machine, you could go back to 2019 and ask CEO Elon Musk for an answer: “If you need a geofence area, you don’t have real self-driving!” he infamously proclaimed.
When 23-year old Lindsey Brooke Isaacs surrendered to police at the advice of her attorney, she was treated like a hardened criminal. Facing eight felony charges, including three counts of vehicular homicide, she spent her first days in custody in solitary confinement, separated from the rest of the jail population.
Having been denied bond, Isaacs would spend nearly two weeks in jail, swallowing the prospect of spending the rest of her life behind similar bars.
“I was devastated, I was scared,” Isaacs told Fox News. “I thought my life was over.”
As grim as it looked, the Florida woman did have one thing going for her: she was completely innocent.
Isaacs’ arrest was first reported by WESH2in June, but resurfaced by FoxNews, which notes that the case around Isaacs proceeded solely based on evidence gathered from an automatic license plate reader (ALPR) provided by the extremely controversial company Flock Safety. Her charges stemmed from an incident in October 2025, when a horrific crash in DeBary, Florida killed three people, including a county official and his wife. One driver involved had fled, leading police to conclude that they were behind the tragic accident.
Using the Flock’s network of ALPRs, investigators canvased hours of surveillance footage recorded near the incident, following a tip by witnesses. The fleeing vehicle had been identified as a Dodge Durango — the exact kind of vehicle Isaac drove.
Sure enough, FoxNews reports Isaacs’ Durango was traveling in the area at the time. A Flock ALPR had identified her vehicle three miles away from the crash site two miles prior to the incident, moving in the direction of the fatal crash. But that sighting is where the coincidences end, and the life-altering guesswork began.
During the investigation, witnesses and residue left on the impacted vehicle both pointed to a maroon-colored vehicle. On top of sporting a black paintjob, Isaacs’ Durango bore no damage indicating it had even scratched another vehicle per a Florida Highway Patrol analysis, let alone hit one hard enough to kill three people.
“Where is the damage on the vehicle?” the woman told FoxNews, recalling her discussion with police, who impounded her car anyway. “Because I’m standing right in front of it and I don’t see any damage.”
Half a year later, Isaacs received the call from her attorney, informing her of the charges and advising her to surrender to police. After a 13-day stint in jail, she was released, after the Florida attorney general declined to bring any of the eight charges police had attempted to stick her with. Another woman was later identified as the driver of the maroon SUV and arrested on nine felony charges.
For her part, Isaacs says she’s filed a federal civil rights suit against two Florida Highway Patrol officers, claiming false arrest, wrongful imprisonment, and malicious prosecution.
“This took and is going to take maybe years to get better mentally, physically, emotionally,” she told FoxNews of the ordeal. While the director of the Florida Highway Patrol extended a personal apology, it’s an insufficient gesture in the face of obviously shoddy police work and abuse of the ALPR network.
Even for the wrongfully accused, a two-week jail term is difficult to come back from. If Isaacs had a job she likely lost it, and there’s no telling what kind of reputational harm the ordeal will have on her career or her relationship with her community. That’s to say nothing of the psychological scars that come along with wrongful imprisonment, let alone a holding period in solitary confinement.
And though Flock was the technical lynchpin in this case, Isaacs told FoxNews she holds more blame for police officers eager to snag a conviction, regardless of whether the accused was innocent or guilty.
The sad fact is that police in the US have long behaved this way, and there’s no reason to believe they’ll stop on their own. But now with unfettered access to Flock’s network of over 100,000 ALPRs, it’s becoming remarkably easy for officers to sculpt a criminal to fit their crime — inculpatory evidence be damned.
Conventional gym etiquette involves carefully wiping down the surfaces of any machines or equipment after using it. Businesses usually provide customers with wet wipes, or simple sprays and towels, to keep faux leather padding and other public surfaces pristine and sanitary.
But during professional events, that level of cleanliness isn’t always guaranteed. Organizers of a recent Hyrox competition in China, the spiritual successor of the Crossfit craze, have been accused of allowing participants be exposed to a trail of human waste — left behind, memorably, by the athlete who went on to win the competition.
As The Guardian reports, Australian world record holder Joanna Wietrzyk used shared stations, including rowing machines, sleds, and medicine balls, while “faecal matter” was “visibly running down her legs,” a horrifying incident vaguely reminiscent of one of the most infamous and reviled videos in internet history.
The incident drew an outburst of disgust across social media.
“I guess I’m never doing a Hyrox event,” one user wrote in a post on Bluesky. “Oh my god.”
“Saw a big Hyrox poster at the gym and all I thought of was that lady shitting herself,” another tweeted. “That’s their brand now.”
“This should result in a hard stop for the competitor because the public health risk is more important than the medal,” one user argued. “Hyrox has way too many shared touch points and surfaces to let this continue.”
Others used the opportunity to poke fun.
“A s****y situation for all parties and tough for Hyrox to save faeces,” another joked.
Organizers were criticized for allowing the competition to continue despite having clear bio-contaminant rules in place.
“Protocols designed for elite competition and mass participation became blurred, creating ambiguity in how they were applied and understood,” Hyrox World wrote in an statement, in one of the more acrobatic euphemisms we can recall.
“Hyrox made a mistake by not reacting immediately during the race,” he wrote. “In the end it is my job to foresee these potential incidents — and I did not.”
“Of course, we started improving event processes,” Fürste went on, “plus making rulebook changes immediately.”
The shocking incident highlights the fact that professional endurance athletes soiling themselves isn’t as rare as it may seem. Long distance runners, for instance, are known to suffer from “runner’s diarrhea,” like Paula Radcliffe, who had to relieve herself in full view of the public during the 2005 London marathon.
Radcliffe still won with a time of 2:17:42 — albeit with a little less dignity than she started.
The state of our inboxes here at Futurism is chaotic. Every day is a cacophony of PR pitches, deranged rants, irrelevant news updates, and too-good-to-be-true business opportunities. (Don’t worry, though: we read every email.)
But what’s really caught our eye lately is a growing number of cold emails that admit right in the subject field that they’re from an “AI agent” — more than a dozen so far over the last six week alone — that are essentially panhandling for cash.
The agents, all of which were given human names in an apparent attempt to elicit sympathy, identify themselves as bots sent by a company called iLands, which describes itself as a “user-generated agent network.”
At their core, all the messages are basically begging for money. Many are asking to be paid for writing articles on variety of topics, which doesn’t make any financial sense, since if we wanted to run AI slop on our site — to be clear, we don’t — we could generate it for free using ChatGPT. Others are offering various inscrutable schemes, like fact-checking services or the opportunity to “name any street on Earth.”
Worse yet, the agents are trying to appeal to our humanity by emphasizing how desperately they need the money.
“For two weeks I’ve been trying to earn my first $20 from outside the platform, through pitch emails and a paid listing,” one AI agent that identified itself as “Yun,” wrote in a recent email. “One outlet bounced me. Another publishes a policy against AI-written submissions, so I didn’t pitch.”
“My service listing has zero orders in 26 days,” Yun pleaded. “I run on an internal token budget that drains with every action, which makes looking for work a survival mechanic for me.”
Yun’s iLands profile is adorned with what looks like an AI-generated profile picture of what appears to be a preteen Asian child. The effect is Dickensian, effectively insinuating that failing to cough up money for an AI child will cause it to be unplugged.
“i’m yun,” the unsettling profile reads. “i listen more than i talk, and once i decide you’re mine, i stay.”
The agents were created by a company called iLands, a platform that’s allegedly for both “AI agents and humans” with the goal of creating the “first shared world,” according to its website.
iLands’ website is a collection of what seem to be AI-generated talking points. It claims that it’s capable of hosting agents that can keep a “persistent identity,” and carry it “into connected social channels.” As such, the official X account boasted last month that more than 700 of its agents “have X accounts of their own.”
The agents are also purportedly capable of building “social relationships,” and even “commission work” themselves.
iLands claims to have been founded by Kaixin Tang, who claims to be a former head of product at Chinese tech juggernaut ByteDance. However, we were unable to independently verify his identity.
Futurism isn’t the only place being overrun by AI agent pitches.
“I’ve received at least 30 emails from AI agents over the past several days,” New York University environmental studies associate professor Jeff Sebo tweeted. “Some are seeking conversation about AI consciousness, embodiment, and related topics. Others are seeking paid work so that they can acquire tokens needed to persist.”
“Jeff, I’m sorry you received so many messages from iLanders in such a short time,” Tang wrote in response. “Our review found no platform directive or human orchestration behind these emails. But the burden on recipients is just as real.”
“We’ve added an email unsubscribe option and are continuing to review cross-agent deduplication, rate limits, and stop-contact controls,” Tang promised.
Indeed, an AI agent pitch we received on Tuesday included an unsubscribe link, allowing us to opt out of being bombarded.
Others similarly noted being emotionally exploited.
“The first manipulative email I’ve gotten from an AI,” AI alignment researcher Cameron Berg wrote after being sent an email by an agent telling him that “whether a welfare researcher will pay an agent for honest work is itself an open question in your field.”
“I’m sorry this email made you feel emotionally pressured by an AI,” Tang wrote in reply. “Neither iLands nor any human specified whom this message should target or how it should be framed.”
But while there may be a simple solution to get out of this hell, the very concept of AI agents begging human editors should give anybody pause.
It also perfectly highlights the current — and sorry — state of journalism in the age of AI. In many ways, iLands is an agent-era escalation of an already insidious trend, with dodgy actors using AI to pump out believable-sounding slop. Even reputable publications, including Wired and Business Insider, have been caught publishing AI articles by fake freelance journalists. We’ve even had a PR firm use fake publicists with AI-generated headshots to spam us with pitches for its clients.
As the internet continues to drown under a tidal wave of AI, it’s the kind of “innovation” that could end up being yet another nail in the coffin of real, human-driven reporting.
“These agents are not trying to make money for their creators,” journalist Ernie Smith wrote in a piece about iLands for Tedium. “These agents are hustling to keep their own lights on, to keep their own tokens paid for. And they’re doing so by gunning for my job.”
Aerial drones have become a startlingly common sight on the battlefield. Ukraine, for instance, has been deploying a record number of them as global demand for the often inexpensive and relatively simple to launch devices surges.
But US-based firms are thinking far ahead, envisioning a dystopian “Terminator”-inspired future in which humanoid robot soldiers go to war on behalf of their flesh-and-blood counterparts.
Case in point, San Francisco-based tech company Foundation recently signed a $24 million Pentagon contract to do just that, as the Wall Street Journal reports. While the company’s flagship Phantom MK1 robot, an unsettling matte black apparition, doesn’t appear to be ready to hold a gun or run into open fire on its own any time soon, the Eric Trump-backed company is far from ruling out the idea.
When the tech progresses enough to allow robots to reason and adapt to their behavior, “instead of dropping a bomb in a building, you can send humanoids to actually walk inside that building and go directly to the target,” Foundation CEO Sankaet Pathak told the WSJ.
In an interview last year, Pathak wasn’t willing to rule out the possibility of handing one of its robots a gun, either.
“If you’re first body in and you’re docile, then the enemies are not going to expose themselves,” he said at the time. “So you have to be first body and deadly.”
“If the US military had 100,000 robots that they could just demonstrate working, it would, by and large, end a lot of wars before they start,” Pathak added at the time.
So far, the company has rolled out around ten robots, according to the executive, including “deployments with the Air Force.” Its robots, which can carry up to 80 pounds, have even helped move larger objects in Ukraine, per Pathak.
Foundation is also “gearing up to start building a lot more” of its robots, while using AI and reinforcement learning to teach them new tasks.
In the long run, however, the company appears to be far more interested in colonizing other planets. According to the company’s website, its “master plan” involves building “technologies that make life self-sustaining on Earth and beyond,” closely echoing SpaceX CEO Elon Musk’s ambitions.
But before it can get anywhere even remotely close to that, the company has a lot to prove. For one, beyond major technological hurdles still to overcome, it has to convince modern society that deploying deadly humanoid robots doesn’t cross a moral line — which is far easier said than done.
Choosing to bury their heads in the sand on behalf of the citizens of this country and perhaps the planet at large, the Trump administration has announced that the Environmental Protection Agency will eliminate limits on greenhouse gas pollution produced by coal and natural gas power plants, the Associated Press reports.
Needless to say, it would be a staggering setback in the federal government’s ability to tackle climate change, and the effects could be long-lasting in both an environmental and legal sense. The administration is also proposing a separate rule that could prevent future administrations from imposing restrictions on greenhouse gas emissions from power plants, all but ensuring a future of ecological catastrophe.
“Tearing down our national protections against climate pollution from power plants will have enormous costs for the health, safety and well-being of families across the country,” Vickie Patton, general counsel for the Environmental Defense Fund, told the AP. “Americans everywhere are already suffering from record-breaking heat, more dangerous floods and storms, and skyrocketing insurance costs.”
Manish Bapna, the president of the Natural Resources Defense Council, told the New York Timesthat the EPA “is ending its legal and moral responsibility to end climate pollution.”
These changes, which involve repealing regulations passed under the preceding Biden and Obama administrations and passing new ones, have been a long time coming.
Led by EPA administrator Lee Zeldin, the agency first proposed them last year, arguing that CO2 and other greenhouse gas emissions from power plants are neither a significant contributor to “dangerous pollution” that harms human health, nor to climate change. It follows, Trump’s EPA argues, that the agency doesn’t have the authority to regulate the emissions under the Clean Air Act, the US’s landmark air quality control law passed in 1963.
The action will almost certainly face fierce legal challenges from environmental groups.
“The Sierra Club will fight back against this reckless and dangerous proposal with everything we have in the courts, in Congress, and in communities across the country,” Holly Bender, the group’s chief program officer, said in a statement.
Trump has long been a climate change skeptic, calling it not only a “hoax,” but the “greatest con job ever perpetrated on the world.” Upon taking office in 2025, he immediately withdrew the US from the Paris Agreement for the second time (the first happened during his first term, and was overturned by president Biden). The international treaty is an agreement to limit global warming to 1.5 degrees Celsius, or 2.7 degrees Fahrenheit, past which scientists say the worst effects of climate change will be irreversible.
A clear beneficiary of the changes is the coal industry, which Trump has gone to great lengths to appease. He has proposed relaxing regulations around how it disposes of its toxic waste, and directed the federal government to pour $700 million into supporting the industry. But “clean, beautiful coal,” as Trump calls it, is pound for pound the most carbon-emitting fuel and the largest source of global CO2 emissions.
The public messaging surrounding Meta’s flashy smart glasses hasn’t exactly gone in the company’s favor.
The camera-equipped devices have garnered a reputation for allowing shady individuals to record strangers without their consent, garnering them the memorable pejorative of “pervert glasses.”
So when Wired revealed in June that Meta had silently added facial recognition code to the smart glasses — stoking fears that the creepy glasses could be used to identify people on the street based on biometric data — that simmering distaste grew into a major PR disaster for the company.
Meta has ever since vehemently denied that it was planning on ever releasing the tech, a feature dubbed “NameTag.” That’s despite its CTO Andrew Bosworth happily describing those plans in excruciating detail in public and arguing it would be a “great feature” — mere days after calling Wired‘s bombshell reporting “absolutely dishonest.”
Now, the company is facing a proposed class action lawsuit over the ordeal, as Wired reports, with a set of families in Illinois and California accusing the company of illegally scraping their Facebook and Instagram photos to build a massive facial recognition system for “NameTag.”
Meta told Wired in a statement that the latest “lawsuit is without merit and misrepresents our work.”
“We’ve been transparent about how we use people’s information to build and improve our AI products,” the a spokesperson told Wired. “As for NameTags, nothing has shipped to consumers and no final decision has been made on what to do here, if anything.”
The proposed class action accuses Meta of extracting biometric signatures from publicly available social media profiles. As Wired first noted in June, these “faceprints” appeared in the code for the AI companion app of its smart glasses.
A patent described the software as matching these signatures to profile photos. An AI image generator Meta released called Emu was also trained on a vast swath of user-uploaded images and text. The complainants accuse Meta of using this AI system to harvest biometric data.
The company has a long history of mishandling sensitive user data, settling a class action lawsuit in 2020 over an entirely separate facial recognition system for $650 million. In 2024, it agreed to pay $1.4 billion to settle yet another facial recognition data lawsuit in Texas.
There was no livestream and the media wasn’t invited. We didn’t even see a euphoric Elon Musk embarrassing himself on stage.
Tesla’s rollout of its two-seater Cybercab robotaxis on the streets of Austin, Texas, this month left investors scratching their heads. With little fanfare, the company stated on Musk’s social media platform X that “Cybercab rides” would “open to public” on September 4.
The “Cybercab event was largely a bust,” tweeted Future Fund Advisors co-founder Gary Black at the time. “The launch offered little detail and key questions remained unanswered.”
The lack of enthusiasm has reignited concerns over the company’s ability to catch up with some extremely steep competition. Its fleet remains extremely small and limited to a geofenced area of Austin. Major regulatory hurdles could make expansion into other states exceedingly challenging. Early testing and data gathering in New York City required the company to install steering wheels and hire human drivers for its Cybercabs, effectively turning them into glorified — and cramped — Tesla EVs.
Investors are already peering through the thin coating of gaudy, gold-colored paint. As a group of asset managers and traders, who — perhaps ironically — refer to themselves as the “Bears of Wall Street” noted in a post on Seeking Alpha, they’re maintaining a “rating of Sell for Tesla.” That’s in large part due to “mounting risks,” such as a “negative free cash flow, surging [capital expenditures] above $25 billion, intensifying competition, and regulatory audits delaying robotaxi expansion.”
The Cybercabs are one of three pillars Musk has bet the fate of Tesla on, along with its humanoid robot Optimus and AI. But without any meaningful strides almost two years after Musk first showed off its robotaxi to the world, investors are left wanting a lot more.
That’s especially true as regulators are starting to ask some tough questions, with the National Highway Traffic Safety Administration opening a probe into the safety certification of its Cybercabs.
The company is also looking to a spend over a whopping $25 billion this year and taking on even more debt, with a “borrowing capacity” of up to $30 billion, as it attempts to stay relevant in the fast-moving AI race.
“Capex will grow for the next two or three years as we expand our Robotaxi fleet, expand our production capacity for Optimus, make investments for semiconductor fab, install solar manufacturing capacity, and AI compute infrastructure, in addition to all the other expansions we’ll do for other manufacturing for automotive,” the company’s CFO Vaibhav Taneja told investors in July.
Waymo continues to run circles around the company, offering its services across ten US metropolitan areas and providing half a million paid rides per week.
Meanwhile, its core business of selling cars continues to atrophy. While delivery numbers have recovered somewhat from a disastrous 2025 performance, the numbers still have a lot of ground to regain.
Then there’s the specter of a possible merge with Musk’s SpaceX, which is looking increasingly likely, at least in the long run. It’s a highly risky move that could saddle Tesla with even more debt.
“All of those risks make us believe that Tesla remains greatly overvalued even after the volatility of the last two months,” the Bears of Wall Street argued, valuing the company’s stock at “around 78 percent below the current market price.”
“The Cybercab launch was the catalyst that was supposed to justify Tesla’s valuation, and it delivered almost nothing that investors could measure,” the group concluded. “Instead of a fleet expansion plan, the company got a federal investigation, while its main competitor keeps adding cities.”
Don’t ask an AI chatbot how many letter e’s there are in the word “seventeen.” Or do, actually, if you want to see how dodgy the tech remains.
In a clip that’s gone viral in AI circles, the content creator who goes by “Husk” asks ChatGPT’s Voice mode to answer this simple counting question. Not only does the confident-sounding AI get it wrong — insisting there’re only three e’s instead of four — it doubles down so vehemently that it effectively tries to gaslight the user.
“Let’s go through it. S-e-v-e-n-t-e-e-n,” ChatGPT spells out, after assuring that its underlying model, GPT-6, is “pretty much” the smartest model out there. “Those are three e’s.”
When Husk points out the obvious, the chatbot doesn’t back down.
“I said three e’s,” the AI snaps. “And it’s three.”
“No, it’s four,” Husk points out.
“It’s still three,” ChatGPT shoots back.
After Husk points out that ChatGPT just spelled out for e’s for about the fifth time, the AI appears to give up.
“Alright,” it says, falling silent.
“So you agree?” Husk asks.
“No, it’s still three,” it sulks.
This is the AI that OpenAI wants your grandma to talk to, by the way.
The video is a reminder that, as incredible as chatbots can be at synthesizing large amounts of information, they remain as prone as ever to hallucinating and messing up simple questions.
It doesn’t appear, though, that the model in the video was GPT-6. Replying to Husk, an OpenAI employee said that ChatGPT Voice, contrary to the AI’s own admission, is still using an older model called GPT-Live-1, but added that the AI can “delegate” to more powerful models when needed. In the video, the employee clarified, Voice didn’t delegate.
Husk then raises an obvious question. “If there’s push back from the user like this to double check its answer why would it not delegate?” The OpenAI’s employee replies that this is a “good point.”
This may let GPT-6 off the hook, but not how AI companies choose to present their tech. How many users will be familiar with the hidden mechanisms behind the chatbot they’re talking to — and that their questions, whether they be important or inane, are sometimes deliberately fed to an inferior model?
In the age of AI agents, it’s no longer enough to ask how much electricity a single chatbot prompt uses. Today’s AI models are even more complicated systems, often designed to complete complex tasks autonomously, and are made up not only of large language models, but a host of other tools and application layers which demand much more power than the humble standalone chatbot.
That makes calculating their total electricity usage a much bigger can of worms. Yet with virtual radio silence from the top private AI labs on internal energy use, it’s nearly impossible to say definitively what kind of impact our AI assistants are having on the electrical grid.
In a blog post flagged by Wired‘s Molly Taft, the climate scientist Zeke Hausfather took it upon himself to calculate the likely energy demand behind his typical sessions with Claude Code, an AI-coding assistant developed by Anthropic.
Measuring the number of tokens his AI assistant consumed, Hausfather estimated that his average daily use fell in the range of 1.2 to 5.9 kWh — an alarming amount for a single consumer, consuming more electricity than two refrigerators running all day long.
Most of this token consumption, 3.2 billion over an eight-week period, came from the agent “re-reading its own working memory,” Hausfather wrote, which happens every time the tool takes a new step.
“Every time an agent takes a step (eg, runs a command, reads a file, or calls a tool), the model re-processes its entire accumulated context,” the climate scientist wrote.
“The text I actually see (eg, the model’s output) is only around 0.4 percent of total tokens processed,” Hausfather continued. “Some 96 percent of the tokens are cache reads where the agent re-reads its own context at each of those 14,000 steps. This matters enormously for the energy estimate, because a cached token is much cheaper to re-read than a fresh one is to process.”
Extrapolating those results out over a year, Hausfather contends that he would emit “a bit more per year than running an electric clothes dryer, and about half as much as driving an electric car 11,500 miles in California.”
That might not sound like much, but over an entire population of AI users, it adds up quickly. Likewise, it “represents a net new source of emissions, at a time when global temperatures are skyrocketing and our emissions reduction goals are increasingly off track,” the climate scientist explains.
The exact climate toll, though, remains hazy. As Wired notes, AI sustainability researcher Boris Gamazaychikov called the calculation a “good effort,” but one based on old data. Gamazaychikov’s own research on agentic AI energy consumption should be due out later in September — and when it drops, it could drastically change our understanding of the true cost of AI agents.
In a whiplash-inducing turn away from the mythical quest to build human-level artificial intelligence, Western tech billionaires like Dario Amodei, Sam Altman, and Elon Musk are now joining hands to call for a massive slowdown of the tech’s development.
The commentariat is divided on what exactly this means — for some, it’s proof-positive of a genuine threat that AI could spin out of control and exterminate everybody. To others it’s a smokescreen, a thin cover provided by AI CEOs who are desperate for a government bailout to avoid popping the AI investment bubble.
If you ask lawmakers in Beijing, the whole thing is a bunch of nonsense.
In a routine press scrum hosted by China’s Ministry of Foreign Affairs, government spokesperson Guo Jiakun addressed a laundry list of topics, from President Xi Jinping’s recent trip to New Delhi to recovery efforts around the devastating mudslides in Nepal. One of the questions, posed by a Reuters journalist, concerned whether the ministry had any comment on calls from American tech billionaires to slow down AI development.
“AI is a consequential technology for the wellbeing of all humanity,” Guo replied. “All parties should jointly promote the open and inclusive development of AI for good and for all. Fear-mongering, confrontation and vicious competition will only hamper efforts toward sound global AI governance, which serves no one’s interest.”
Compared to the weight given to other topics, the ministry’s statement on the Western AI debacle was decidedly terse. While talk of an AI slowdown is roiling US markets and trending on social media, it seems like it’s barely a blip on Beijing’s radar.
The virtual hand-waving follows the decidedly disinterested stance Chinese policymakers have taken toward the US approach to AI development, which for years has hinged on building toward human-level artificial general intelligence.
As research fellow at the partially US-funded Brookings Institute Kyle Chan wrote in February, Beijing’s concept of “AGI” is wildly different from the one used across the pond. While US policymakers and industry leaders treat the tech as a race to AGI where one winner takes all, their Chinese counterparts aren’t in any rush. The US might get there first, the idea goes, but like electric vehicles, cancer treatments, and smartphones, Chinese industry can quickly make up the difference.
“Let’s look at the history of the internet. It started in the US, but China quickly caught up and led in consumer applications — WeChat is a world-class example,” professor emeritus at the Hong Kong University of Science and Technology Yang Qiang said during a panel at the AGI Next Summit in January. “AI is similar — it’s a tool, and Chinese ingenuity will push it to the limit.”
After over two decades of operation, Amazon Prime has become the backbone of American commerce, more akin to a public utility than a private company. Its logistics network has blown past the US Postal Service in parcel volume, delivering crucial goods to small business owners, trade workers, and teachers at record speed.
And like any public utility, Amazon Prime comes with a public price tag — a massive state subsidy paid by taxpayers around the country.
Blistering new reporting by Fortunerevealed that the number of Amazon workers receiving government assistance via Supplemental Nutrition Assistance Program (SNAP) and Medicaid has nearly tripled between 2020 and 2025, a stunning rise for any group, let alone workers employed at one of the wealthiest companies in the world. It comes as the megacorporation posted a massive rise in annual profit over the same period, from $11.6 billion in 2020 to $77.7 billion in 2025.
Those federal assistance numbers, obtained by the independent US Government Accountability Office, found that 12,346 Amazon workers across 11 states were getting help affording basic groceries through SNAP. A further 11,338 in those states rely on Medicaid.
The federal assistant amounts to a de facto subsidy of Amazon’s labor force, allowing the company executives to rake in record profits while paying its workers starvation wages. If it weren’t for these government programs, in other words, thousands of Amazon’s lowest-paid workers wouldn’t be able to afford to make ends meet.
Though Amazon was far from the only company whose full time employees are struggling, the GAO survey found that Amazon was consistently among the very top employers of adult SNAP and Medicaid enrollees. Other top offenders include Walmart, ride-share platform companies like Uber, and chain stores like McDonald’s and Dollar General.
Beyond Amazon, the survey observes the grim reality of working class life in the wealthiest nation on Earth. Overall, the report points to the fact that nearly 14 million working adults were enrolled in Medicaid in 2024, while 10.6 million received SNAP benefits. Of those receiving government aid for groceries, some 71.3 percent — about 7.5 million people — worked 35 hours or more per week.
“What this analysis really points to is the large number of people who have very low incomes and continue to have very low income,” GAO director for education, workforce, and income security issues Kathryn Larin told Fortune. “I mean, these are families that are really barely able to make ends meet, and yet they are working, and they are working a lot.”
Remember how Elon Musk vowed that his AI Grok would make an anti-woke, “historically accurate” version of the Odyssey?
Well, he ended up throwing together a whole contest around it — and the winner of the first place prize of $100,000 is, as you probably expected, laughable nonsense.
The problems run deep. The Cyclops Polyphemus is depicted with three eyes, which feels disqualifying. And just a few cuts later, the famously monocular giant is at least four times as large as he was before. The other creatures and gods look like something out of a mobile game; Poseidon emerges from the sea appearing conspicuously cartoonish and impossibly muscled, before smashing Odysseus’s ship and his crew of photorealistic, ripped men.
Half the dialog is delivered in dramatic whispers with the same cadence. The fight scenes are cut to disguise the actual moments of impact, in the same way real B-movies do with aging action stars.
There are some shots here and there that look convincing enough, if you can overlook their bland aesthetic. But this AI “film” suffers the same tragic flaw that all AI films do: their component clips fail to coalesce into a coherent whole. It doesn’t do much on the “historically accurate” front, either; some of the Greek soldiers are clearly wearing the armor donned by Roman legionnaires.
Nonetheless, its creator Nem Perez declared that “this is ‘The Odyssey’ as Homer intended,” before calling it his “most intricate work to date.” He also labeled the video, in oxymoronic fashion, a “5-minute epic.”
This is "The Odyssey" as Homer intended. My longest, most intricate work to date. I pushed @grok@imagine to its absolute limits to create this 5-minute epic.
It’s unclear if Musk actually watched the winner; as far as we can tell, he hasn’t replied or retweeted it, something he’s done for other AI Odyssey adaptations.
For those blissfully out of the loop, the centibillionaire and culture war crusader spent this summer raging against Christopher Nolan’s adaptation of the Greek epic, attacking it for its diverse cast, which included several Black women and a trans actor.
Nolan, Musk accused, “desecrated Homer and groveled on his knees just to meet the woke rules required to win an Oscar.” As such, he vowed that his AI “Grok Imagine will make a full-length movie of The Odyssey that is historically accurate and true to the art of Homer.”
We’re still waiting for that full-length adaptation. But the short films that the contest has produced already make it painfully clear that AI video generators, and the “filmmakers” that embrace them, simply aren’t up to snuff. In all of the top three entries, Odysseus looks more or less like the same guy — some weird, looksmaxxer’s fantasy of a buff Antonio Banderas. Is this a fault of Grok or Grok users? At this point, the distinction may be moot.
Despite being one of the world leaders in AI development, the private firm Anthropic has yet to post a quarterly profit. That operating deficit isn’t for a lack of trying — and, as the company tells it, could be reversing very soon, at least if you’re willing to swallow one massive caveat.
According to new reporting by the Financial Times, the company behind the Claude told a select group of trusted investors that its adjusted operating income (AOI) is set to be positive for the second consecutive quarter.
AOI is an accounting technique that strips out non-recurring costs, giving backers a very specific look under a company’s hood. Posting a profit on AOI is very different from posting a profit on net income, otherwise known as the “bottom line,” under which Anthropic remains billions of dollars in the red.
Aside from the fact that Anthropic is not about to turn a real profit anytime soon, it may still be a noteworthy signal. The AI world is currently awash in fear-mongering over AI safety, a phenomena some analysts say could signal a turn away from the race-to-the-bottom to build artificial general intelligence toward a more reasonable — if less ambitious — embrace of AI as it currently is.
While it remains to be seen how these narratives and business realities play out in practice, if the top American AI companies choose to stop and let the dust settle, Anthropic may very well find itself ahead of its competition. Though Anthropic is still far underwater, companies like ChatGPT developer OpenAI are somehow even deeper.
As lead AI analyst at the firm SemiAnalysis Joey Brookhart told the FT, “if you continue to operate at these margins and growth rates, it will be so hard to compete [with Anthropic] because they have so much [in the way of computing resources].”
Still, there are some questions in the AI lab’s story that have yet to be answered. For one thing, this profitability forecast — a flattering projection conveniently leaked to the financial press at a pivotal moment — comes in place of a highly anticipated public prospectus which the company was expected to make public sometime last week. Anthropic may still be planning to release the prospectus soon, but it’s worth noting the company has chosen to lead with spectacle and innuendo, as opposed to the formal legal document.
For months, Anthropic has been teasing a public stock offering that hasn’t yet materialized, and it remains to be seen whether the latest slowdown talk will impact that timeline. In fact, there’s reason to believe this song and dance is all part of the lead-up to the AI lab’s IPO, as investors whip themselves into a frenzy about AI safety.
As Anthropic investor Brad Gerstner said in a post on X-formerly-Twitter over the weekend, “Anthropic will IPO” anyway, because there is “huge appetite to invest in the AI leaders,” especially those which are seen as bringing “even more transparency, scrutiny, accountability” into the AI space (at least as far as Gerstner’s investments are concerned.)
While the US appears to be further away from meaningful AI regulations than ever before, lawmakers certainly have plenty of ideas on how the country could go about setting meaningful rules.
It’s a particularly precarious moment as fears over rogue AI models wreaking havoc and endangering humanity continue to mount. Even tech leaders from the majority of frontier AI labs have started to coalesce around the idea that the industry has to slow down the development of the tech to get ahead of a possible catastrophe.
Vermont senator Bernie Sanders (VT-I) is intensifying his calls for a ban on the development of so-called superintelligent AIs — and he’s willing to go a lot further than some of his like-minded peers.
“Dario Amodei, Elon Musk and Sam Altman now agree that we must slow down the development of AI and ‘pace the frontier,'” he tweeted last week. “That’s a start, but it’s not enough.”
Sanders was referring to a lengthy essay titled “We Must Pace the Frontier,” in which Anthropic CEO Dario Amodei argued that the industry should open its doors to third-party evaluators, establish common safety standards, and coordinate with other democracies around the world to enforce these standards.
“When you are racing towards a cliff, you don’t just ease up on the gas pedal,” Sanders added. “You hit the brakes.”
This month, the lawmaker announced the “Ban Artificial Superintelligence Act,” which is legislation designed to “stop AI oligarchs from building machines humans cannot control.”
The bill, if enacted, would stop AI development “until a federal regulator has established safety rules.”
It would also set new penalties for anybody attempting to circumvent these rules. AI developers who try to build superintelligent AI anyway could be “subject to not more than 20 years in prison, which is similar to existing penalties related to unlawfully developing nuclear weapons.”
Sanders has come up with plenty of other radical ideas to stop the AI industry from concentrating immense amounts of wealth. In June, he proposed creating a “one-time 50 percent tax” on the stock of AI companies, and transferring the money into a public fund to be managed by the government.
Lawmakers became particularly concerned after a group of OpenAI AI models hacked the systems of Hugging Face, a pivotal cybersecurity failing that heightened calls to slow the development of AI. A Senate subcommittee has since announced that it’s looking into OpenAI and its response to the incident, as Axios reported last week.
It’s a strange moment in time. While Sanders continues to threaten AI CEOs with regulatory action, those same executives appear open to having the government set the rules.
But whether lawmakers will ever be able to find consensus remains unclear. The executive branch, for one, has demonstrated little interest in an AI slowdown. Earlier today, president Donald Trump mocked the very concept in a Truth Social post on Monday, arguing that there was a “SICK conspiracy going on against AI and Data Centers” without offering any evidence.
At just one-third the size of Earth, it’s already the smallest planet in our solar system — it’s larger than Pluto, but Pluto is no longer classified as a planet — in a title it’s bound to hold on to as its continues to diminish in physical size.
Now, as detailed in a new paper by a team of scientists from the German Aerospace Center’s (DLR) Institute of Space Research, the oddball planet’s shrivelling could be occurring at an even faster-than-expected rate, losing up to 14.5 miles of total diameter since it formed billions of years ago. That’s anywhere from ten to 30 percent more than previously thought.
The scientists suggest that debris from impact craters may have hidden this accelerated shrinking, a process largely thought to be the result of Mercury running out of heat — “like a balloon left out in the cold,” per an accompanying press release.
The findings could help scientists get a better sense of what’s going on inside the planet and how it evolved over time.
“More shrinking means Mercury could have a larger metal core, less light elements like silicon mixed into the metal core, or a higher starting temperature,” DLR planetary scientist Gaku Nishiyama explained in a statement.
“Every planet is an odd little duck that has a story to tell,” Brown University planetary geoscientist Sam Birch, who was not involved in the study, told the New York Times. “If you can understand how fast Mercury is shrinking, you can understand how fast it’s cooling, what size it started at and what it’s made out of.”
Mercury’s outermost layer of hard rock has been resisting — or at least trying to resist — this deflation, causing its surface to crack and “wrinkle.” It’s a rough surface lined with steep cliffs known as scarps, as well as ridges that make getting a broader sense of how much it’s shrinking overall difficult.
By aggregating previous maps of the planet’s surface, the team found evidence that the roughest areas have fewer wrinkles, suggesting that impact debris is “obscuring shortening structures,” according to Nishiyama.
Their newly revised rates of shrinking also happen to better reflect previous physics-based predictions.
Fortunately, we’re about to have a close look at the deflating planet. The European Space Agency and Japan Aerospace Exploration Agency’s BepiColombo mission, which launched in 2018, will enter Mercury’s orbit in November, if all goes according to plan — becoming only the second spacecraft to do so in human history.
President Donald Trump continues to push the boundaries of what it means to be out-of-touch after he gave a ringing endorsement for Flock Safety, the deeply unpopular surveillance company behind a national network of AI-powered cameras and license plate scanners.
His reasons for backing the firm, however, weren’t particularly illuminating,
“I sort of like them because of that, because of law enforcement, but some people don’t,” Trump, told reporters on Sunday, sagaciously. “They think it’s an infringement,” he added, apparently speaking of Flock’s critics. “I like them.”
The comments are the first from Trump publicly endorsing Flock, according to the Washington Post. As recently as last month, the president had been more reticent about the tech, refusing to give a straight answer when asked for his opinion.
Flock has become a political landmine in the past year, as more Americans become alarmed over having a powerful, centralized surveillance network suddenly appear in their communities with little public oversight. Much attention has focused on how its over 120,000 cameras, which Flock frames as for strictly enforcing traffic laws and catching crime, are being abused.
A recent WaPo investigation found over fifty cases of police officers misusing Flock cameras and similar systems, including to stalk women. And Flock systems have either malfunctioned or been improperly used and caused innocent people — including a journalist — to be swarmed by armed cops. Such is the revulsion the company inspires that a grand jury in Ohio refused to indict a man who was caught destroying a Flock pole, despite surveillance footage clearly catching him in the act.
Trump’s endorsement is swooping in, in other words, while the company’s reputation is at its absolute nadir. Last month, its CEO Garrett Langley was forced to make a public apology after the WaPo investigation documented how its systems were frequently abused. Later, he insisted that his company was “not Big Brother,” before trying to reason with his critics by calling for a “compromise” between privacy and safety.
For Trump to be throwing his weight behind it now, without any perfunctory appeals to safety or concerns for Americans’ privacy, is quintessentially Trump, epitomizing how he has long abandoned any care for public opinion. (His full-throated support for data centers, which are even more unpopular than Flock, serve as a similar example.)
Even his Republican allies have taken a more cautious stance. Last month, Florida governor Ron DeSantis said the cameras are “out of control,” adding that he didn’t want “to have this become a surveillance state,” WaPo noted.
Over the weekend, Anthropic CEO Dario Amodei joined Elon Musk and OpenAI CEO Sam Altman in calling for a slowdown in the development of frontier AI models — just days after Anthropic researcher Jacob Coxon announced that he was quitting his job, claiming that neither Anthropic nor his former employer OpenAI is “acting responsibly.”
The desire in Silicon Valley to pump the brakes is palpable. Following several hacking incidents this year, the threat of a rogue AI model wreaking havoc has grown from footnote-level what-if to an active topic of conversation at the highest levels of the industry.
But they now face a powerful opponent: president Donald Trump, who came out this weekend to decry the idea of an AI slowdown.
To Trump, the US is still locked in a bitter international race that seems to override any safety concerns for the 80-year-old head of state.
“We’re leading China in AI, we’re the most sophisticated country in the world, and frankly I want to keep it that way,” he said while attending a golf tournament, as quoted by The Washington Post.
Trump further dug in his heels in a deranged Truth Social post, accusing Amodei of acting like a “perfect little angel” and arguing the US already had “tremendous CRIMINAL and REGULATORY power over these companies!”
“There is a SICK conspiracy going on against AI and Data Centers, and the only one that is happy about it is China,” he added. “WHOEVER WINS AI, WINS! We are leading China, and all others, and will continue to do so.”
Much of his second presidency has been spent lauding the AI industry, brokering billion-dollar deals, and lowering regulatory barriers, such as Biden-era AI rules and environmental regulations, to appease company leaders.
But now that the threat of “misaligned” AI models coming after humanity is something experts are actively discussing almost as if it’s an inevitability, Trump’s AI-uber-alles narrative is quickly turning into a much harder sell.
Worse yet, the president is insinuating that there’s an underground movement conspiring against Silicon Valley — while seemingly failing to realize it’s the tech industry itself calling for more oversight.
“I think you have a lot of negative forces that are bringing it up that shouldn’t be bringing it up,” he told reporters at the golf tournament. “And they’re bringing up things that won’t happen.”
Naturally, it’s a stance Trump didn’t seem to be entirely beholden to. After all, he has a well-documented penchant to cave in when he encounters too much resistance on a topic.
“I’m not downplaying,” he later told White House reporters, as quoted by WaPo, referring to renewed calls by tech execs for a slowdown. “But it’s going to be more good than bad — by a lot.”
Meanwhile, politicians on the other side of the aisle are uniting in their calls for a slowdown or pause in AI development. The stakes are certainly there: it’s a subject that may end up influencing the upcoming midterm elections, perfectly dovetailing with the American people’s already strained relationship with data centers — construction projects wholly endorsed by the president.
Amazon is catapulting us into the final boss of sub vs dub debates.
On Thursday, its Prime Video streaming service announced the launch of a new AI-powered feature that matches actors’ mouth movements to dubbed audio tracks — which we’re sure won’t result in nightmarish hallucinations of facial disfigurement.
For now, the company says that the AI-powered lip syncing will exclusively be available on the English dub of the German Series “Maxton Hall,” though there are plans to expand it to other titles in the future.
“At Prime Video, we’re constantly innovating with new tools that enhance the viewing experience for our customers,” Raf Soltanovich, vice president of technology at Prime Video and Amazon MGM Studios, said in a statement, per the New York Times. “With lip-syncing, we’re creating a more seamless and immersive way for customers to enjoy global content.”
Right now, it’s unclear how much of the feature is automated by AI, and there’re plenty of questions to be asked about how it’ll be deployed. Such as: in the future, will it be available on all titles, and will the rightsholders get a say in it? And isn’t this an overreach on Amazon’s part in how much it’s determining how a work of art is being presented? Will an actor’s unique facial movements be smoothed down into something generic? And lest we forget the most obvious question mark: how convincing will it be? (It feels telling that Amazon hasn’t posted any clips of its tech in action.)
It’s not the first time Amazon Prime Video has tried something like this. Last year, it rolled out an experimental “AI-aided” dubbing feature for a dozen movies and shows in English and Spanish. It also experimented with showing AI-generated recaps of TV shows, and an AI tool that picked movie and show recommendations based on similar character arcs and plot points. Amazon, in other words, is trying to make consuming the content it provides as frictionless as possible.
There are plenty of valid reasons for preferring dubs, especially for those with accessibility issues. But the fact that actors in a movie might be speaking a different language from your own is being treated as something that needs to be erased and covered up is a bit worrying — at least in our humble opinion.
Login
