Like any crappy human writer, AI chatbots have a tendency to overuse specific words — and now, scientists are using that propensity to catch their colleagues when they secretly use it in their work.
As the <a href="https://www.nytimes.com/2025/07/02/health/ai-chatgpt-research-papers.html">New York Times reports,</a> scientists estimate, based on an analysis of those overused terms, that there could already be hundreds of thousands of academic papers written with the assistance of AI.
In a new study published in the <a href="https://www.science.org/doi/10.1126/sciadv.adt3813">journal Science Advances</a>, researchers from Germany's University of Tübingen identified some 454 words that large language models (LLMs) use frequently — terms like "garnered," "encompassing," and "burgeoning" — and found that anywhere from 13.5 to 40 percent of biomedical article abstracts were written entirely or with assistance from AI.
With roughly 1.5 million papers indexed each year on the academic journal database PubMed, that means that at least 200,000 of those papers could have been written with the help of LLMs. That whopping figure may, as the NYT notes, be conservative when accounting for any intentional editing of AI-generated text.
While some deceptive journal writers take pains to conceal their AI use, others didn't seem to care who knew. In an <a href="https://x.com/rao2z/status/1769429746717515853">example posted on X</a> by Arizona State University computer scientist Subbarao Kambhampati, for instance, the "writers" of a low-quality radiology journal left in an acknowledgement that it had been penned by a chatbot.
"I'm very sorry," the paper quote reads, "but I don't have access to real time-information or patient-specific data as I am an AI language model."
Not all mistakes are going to be that obvious. Unless you're familiar with the term "regenerate response" — an option on ChatGPT that forces the chatbot to rework a shoddy answer — you could easily miss it being sprinkled throughout respected journals, the blog <a href="https://retractionwatch.com/2023/10/06/signs-of-undeclared-chatgpt-use-in-papers-mounting/">Retraction Watch found</a> back in 2013.
That same blog also flagged a frustrating debacle involving a paper about millipedes with completely made-up references, which was <a href="https://futurism.com/professor-chatgpt-scientific-paper-errors">initially withdrawn</a> from a pre-print server, only to reappear online on a different academic database with those same hallucinated sources.
And let's not forget the journal that was <a href="https://www.vice.com/en/article/scientific-journal-frontiers-publishes-ai-generated-rat-with-gigantic-penis-in-worrying-incident/">forced to retract a paper</a> after observers noticed it was filled with nonsense, including an AI-generated image of a rat with comically gigantic genitals.
Complicating matters, academics are taking pains to hide their AI use. As Kambhampati told the NYT, academics have even started changing the way they write so their work won't be mistaken for AI, removing terms like "delve" that are overused by LLMs.
In the Science Advances paper, the Tübingen researchers proffer that all this AI use in academic journals, if corroborated, could be an "unprecedented impact on scientific writing in biomedical research, surpassing the effect of major world events such as the COVID pandemic."
To the mind of coauthor Dmitry Kobak, however, the whole thing seems counterintuitive.
"I would think for something as important as writing an abstract of your paper," he told the NYT, "you would not do that."
More on AI: <a href="https://futurism.com/openai-forensic-psychiatrist">OpenAI Says It's Hired a Forensic Psychiatrist as Its Users Keep Sliding Into Mental Health Crises</a>

Paper Planes

Astonishing.

Bombshell Research Finds a Staggering Number of Scientific Papers Were AI-Generated

In 2025, people are finding all kinds of uses for artificial intelligence. Look no further than the world's chief contributors to pollution, for whom the buzzy new tech offers bold new possibilities to spew corporate propaganda.
Louis Anthony "Tony" Cox Jr is a petrochemical industry mouthpiece and <a href="https://www.latimes.com/local/california/la-me-epa-pollution-study-20190321-story.html">former Trump advisor</a> who's now hard at work developing a large language model (LLM) program to censor tree-hugger "propaganda" from epidemiological research, <a href="https://www.theguardian.com/technology/ng-interactive/2025/jun/27/tony-cox-epidemiology-risk-assessment-chatgpt-ai">according to the Guardian</a>.
Cox — who <a href="https://www.motherjones.com/environment/2018/10/tony-cox-trumps-air-pollution-adviser-clean-air-saves-no-lives/">once claimed</a> there's no connection between air pollution and respiratory problems — has a long history of collaboration with such model citizens as tobacco monopoly <a href="https://www.jstor.org/stable/10.1086/653091">Philip Morris USA</a>, the <a href="https://jacobin.com/2025/02/chemical-industry-zeldin-trump-lobbying">American Chemistry Council</a>, and the <a href="https://www.capradio.org/news/npr/story?storyid=1141303693">Truck and Engine Manufacturers Association</a>. Cox's relationship with the American Petroleum Institute, the Guardian notes, is so cozy that he's even let the petrochemical lobbying group "<a href="https://subscriber.politicopro.com/article/eenews/1060109129">edit</a>" his research.
He began work on his LLM back in 2023, after a "long experimental chat" with ChatGPT, in which he argued about pollutants and asked it to generate sonnets about Abraham Lincoln and <a href="https://www.epa.gov/pm-pollution/health-and-environmental-effects-particulate-matter-pm">PM2.5</a>, a type of toxic particulate matter. The latter query, the Guardian highlights, led the line "no nose can catch you, for you are so fine."
Dissatisfied with the length of time it took him to convince the <a href="https://futurism.com/chatgpt-users-delusions">notoriously agreeable</a> ChatGPT that PM2.5s don't cause lung cancer — a career-long <a href="https://www.sciencedirect.com/science/article/abs/pii/S0013935123001032?via%3Dihub">hobby horse</a> of his — Cox turned to his buddies in the petrochemical industry to pitch an app that could be pre-trained to avoid such inconvenient facts. Or, as he put it, to do "critical thinking at scale."
That app, of course, is one which will tell users what Cox and his friends of industry want them to hear. His painfully obvious track record, including studies like a chemical industry-sponsored report on the supposed <a href="https://pubmed.ncbi.nlm.nih.gov/38832926/">safety of PFOAs</a>, will no-doubt form the foundation of his LLM's training.
Like <a href="https://compass.onlinelibrary.wiley.com/doi/10.1111/gec3.12736">greenwashing</a>, the corporate practice of misrepresenting something as environmentally friendly, experts say Cox's "AI-washing" is a perfidious form of mass-market propaganda, meant to distance seedy characters like himself from the stigma that "petrochemical lapdog" carries.
"Instead of having scientists-for-hire do that denial work, which advances their economic interests, the industry is funding efforts to outsource it to a machine in order to give it an image of unbiased neutrality," Itai Vardi of the Energy and Policy Institute told the Guardian.
"AI language models are not programmed, but built and trained," Vardi continued, "and when in the hands and funding of this industry, can be dangerous as they will further erode public trust and understanding of this crucial science."
Cox's insistence on AI-washing public health on behalf of his corporate masters is doubly ironic given AI's <a href="https://futurism.com/ai-pollution-carbon-energy">substantial impact</a> on the environment — especially through air pollution. Researchers have recently found links between the <a href="https://futurism.com/cancer-ai-data-centers">coal-powered generators</a> used to power AI datacenters in the United States and a drastic uptick in premature cancer and asthma deaths.
If, or perhaps when, Cox's big business bot comes online, it probably won't make a big scene. Instead, it'll likely seep in unnoticed, part of the unrelenting flood of new AI tools — and all the <a href="https://link.springer.com/chapter/10.1007/978-3-031-73741-1_21">consequences</a> that come with it.
More on Pollution: <a href="https://futurism.com/elon-musk-memphis-illegal-generators">Elon Musk Reportedly Doing Something Horrid to Power His AI Data Center</a>

Yes Man

That's one use for AI.

Pollution Industry Using AI to Make the Case for More Pollution

For years, scientists have <a href="https://phys.org/news/2023-09-supernova.html">suspected</a> that stars can meet their doom by a one-two punch of back-to-back explosions — but they've never seen visual evidence of this happening.
That just changed. Astronomers using the Very Large Telescope in Chile have taken the first-ever image of a star that died in a stellar "double-detonation," leaving behind a spectacular supernova remnant.
Their findings, published as a <a href="https://www.nature.com/articles/s41550-025-02589-5#:~:text=Our%20observations%20reveal%20two%20distinct,mechanism%20could%20occur%20in%20nature.">new study</a> in the journal Nature Astronomy, deepen our understanding of the stellar evolution of burned-out stars called white dwarfs.
"The explosions of white dwarfs play a crucial role in astronomy," lead author Priyam Das, a researcher at the University of New South Wales Canberra, Australia, said in a <a href="https://phys.org/news/2025-07-detonation-image-star-destroyed-pair.html">statement</a> about the work. "Yet, despite their importance, the long-standing puzzle of the exact mechanism triggering their explosion remains unsolved."
Once an exceptionally massive star — one at least several times heavier than the Sun — burns through all its fuel, it collapses under its own gravity in a powerful explosion known as a supernova. 
That's just one way supernovas can happen, though, and not all of them end the same. Some result in the star being completely obliterated, but others, if the star is heavy enough, can produce a super dense core called a neutron star, or even a black hole.
The scene imaged by the VLT is the work of what's called a Type Ia supernova, produced by a low-mass star that exhausted all its fuel and left behind a remnant called a white dwarf. These objects are more compact and far denser than their original stars, endowing them with a wicked gravitational pull.
In binary systems, this powerful gravity can lead to the white dwarf stripping matter off its stellar companion if their orbits are close enough. When enough of this stolen material accumulates on the surface of the white dwarf, reaching a point known as critical mass, it kickstarts a single but incredibly destructive thermonuclear explosion that wipes out both stars.
That's the typical understanding. More recent research, though, has found evidence that some white dwarfs are battered by two explosions, not one, prior to winking out. In this scenario, astronomers believe that a white dwarf is swimming in a cloud of siphoned helium. This unstable helium cloud is the first to explode, precipitating a second blast in the core of the star. And bam: you have a two-fer supernova.
Critically, this type of supernova occurs before the white dwarf reaches critical mass. Astronomers predicted that this double-detonation would produce a unique, visual signature in the form of two separate shells of calcium — and the new image bears this out. If you look closely, you can see that the calcium, depicted in blue, is indeed in a two-shell arrangement.
This is a "clear indication that white dwarfs can explode well before they reach the famous Chandrasekhar mass limit, and that the 'double-detonation' mechanism does indeed occur in nature," said coauthor Ivo Seitenzahl, who conducted the observations while at the Heidelberg Institute for Theoretical Studies in Germany, in the statement.
The work is invaluable for another reason. Type Ia explosions are considered "standard candles" that astronomers use as a measuring stick in the cosmos, because they shine at a consistent luminosity. Now we understand a little more about why that's the case.
"Revealing the inner workings of such a spectacular cosmic explosion is incredibly rewarding," Das said.
More on stars: <a href="https://futurism.com/scientists-signal-early-universe">Scientists Working to Decode Signal From Earliest Years of Universe</a>

Super-duper-nova

This is stunning.

Astronomers Capture First-Ever Image of Star That Exploded Twice

<h2>AI Kirk</h2>
Despite being <a href="https://futurism.com/frontier-ai-models-stumped-childrens-game">stumped by simple children's games</a> and <a href="https://futurism.com/the-byte/openai-new-ai-problem-hallucinate-more">uncontrollably hallucinating</a>, AI models could perform surprisingly well when put in charge of navigating space inside a simulated spacecraft, researchers have found.
As <a href="https://www.livescience.com/space/space-exploration/chatgpt-could-pilot-a-spacecraft-shockingly-well-early-tests-find">Live Science reports</a>, a team of scientists from MIT and the Universidad Politécnica de Madrid, Spain, instructed OpenAI's blockbuster AI chatbot ChatGPT to "operate as an autonomous agent controlling a pursuit spacecraft."
To their amazement, as detailed in <a href="https://arxiv.org/abs/2505.19896">a paper</a> slated to be published in the Journal of Advances in Space Research, they found that the large language model exceeded expectations, scoring second place in a space simulation competition based on the popular video game Kerbal Space Program that pitted several AI agents against each other.
<h2>To Infinity and Beyond</h2>
To test how autonomous agents could be used to maneuver satellites and other space-based assets, researchers created a software design challenge called the <a href="https://www.ll.mit.edu/conferences-events/2024/01/kerbal-space-program-differential-game-challenge">Kerbal Space Program Differential Game Challenge</a>. Participants are encouraged to employ a number of different approaches, such as reinforcement learning.
As part of the program, scientists pitted various autonomous systems against each other in several different scenarios.
In the latest research, the team developed a way to translate instructions and queries about how to orient and maneuver a spacecraft into text and fed it to several commercially available LLMs. With only a small number of prompts, they found that ChatGPT, in particular, performed surprisingly well, coming in second place in the Game Challenge. First place, in an important caveat, went to a system based on equations that model spaceflight.
Best of all, the research was conducted before OpenAI released its GPT-4 model in March 2023, suggesting the latest iterations might perform even better.
But there are still plenty of reasons why we should think twice before handing over the controls to the tech. For one, the current crop of AI models still <a href="https://futurism.com/ai-industry-problem-smarter-hallucinating">suffers from rampant hallucinations</a>, which could make for an awkward — and particularly terrifying — twist in a deep space mission.
More on AI: <a href="https://futurism.com/the-byte/tech-workers-ai-lay-offs">Tech Workers Say They're Rapidly Being Replaced by AI</a>

AI Take the Wheel

It's like Autopilot for space.

Rocket Scientists Hooked Up ChatGPT to the Controls of a Spaceship, and the Results Were Not What You Might Expect

The digital platform era hasn't been kind to musicians, to put it lightly.
Though platforms like Spotify and Apple Music make it easier for audiences to access their favorite bands than ever before, the algorithms and contracts behind those apps are ruthless, paying artists fractions of pennies for their work. Though companies like Spotify have <a href="https://www.theregreview.org/2024/05/30/stern-the-inequalities-of-digital-music-streaming/">tripled their value</a> during the COVID-19 pandemic, the number of artists who are even <a href="https://jaxmusic.org/jme/2023/10/spotify-pay-change/">eligible for royalties</a>, regardless of the paltry amount, continues to shrink.
Now, the rising tide of <a href="https://futurism.com/suno-ai-ceo-making-music">generative AI</a> is boxing out real artists as those same music algorithms become infested with gobs of computer-generated slop.
As <a href="https://pitchfork.com/thepitch/how-ai-wreaked-havoc-on-the-lo-fi-beat-scene/">Pitchfork's Kieran Press-Reynolds</a> explored in his latest column, the "lo-fi" genre — which <a href="https://iaspmjournal.net/index.php/IASPM_Journal/article/view/949">emerged in the 2010s</a> in an earlier, though less overt era of algorithmic chaos — is fast becoming a wasteland of unchecked beatspam thanks to the rise of AI bots.
Younger app dwellers will know lo-fi as the dreamy, unobtrusive genre that first took shape via hours-long beat mixes on YouTube, usually dressed up with nostalgic videos playing on loop. Its devotees make up a close-knit <a href="https://musicfinland.com/en/news/tomppabeats-kupla-and-idealism-superstars-of-low-fidelity-and-low-profile">digital community</a>, sharing skills, advice, and new tracks via social media and messaging apps. At least, they used to.
Now, Press-Reynolds writes, those artists are losing out on royalties and production contracts, as they struggle to discern which tracks were even made by a human, and which have been pumped out by an algorithm. Many are leaving the scene altogether, saying the genre's been "overtaken" by AI crud.
"Previously, you could stream a track on Spotify or Apple and almost be certain you followed them on Instagram or spoke to them on Discord because the community was so tight-knit," a beatmaker named Mia Eden told Press-Reynolds. "Now, it feels so nameless — where this could be an artist that maybe doesn’t like to show face, or it’s a computer. You can’t always distinguish now, and I’d say it’s over half [AI]."
Alex Reade, another artist from the UK, streams his work under the pseudonym <a href="https://youtu.be/iwrjRC2qcXU?feature=shared">Project AER</a>. His once-growing Spotify account previously clocked two million listeners a month; now it's down to less than 500,000.
While Spotify's notoriously opaque <a href="https://dukespace.lib.duke.edu/server/api/core/bitstreams/d7f1bd3e-b4f6-4f57-ba16-4b26383cc892/content">royalty system</a> is hard to parse by design, artists who've similarly reported clocking <a href="https://www.danielinskeep.com/blog/how-much-spotify-pays-me-for-2-million-streams">2 million streams</a> say that's worth a little over $7,600, not accounting for expenses. That's a lot of revenue to dry up overnight — indicative of the <a href="https://futurism.com/fansly-outlaws-furry-content">precarious nature</a> of the platform economy.
"I’m trying to find any other means so I can take that reliance that I have on lo-fi out of my life because it causes me a lot of stress," Reade said.
Lo-fi isn't the only genre being polluted by slop. Earlier this week, social media users flagged an "indie rock band" called <a href="https://futurism.com/ai-band-spotify-streams">The Velvet Sundown</a> which appears to be entirely generated by AI.
The "band," which now boasts over 750,000 monthly listeners, has no presence beyond social media. Everything it puts out — band photos, album covers, and especially the music — show <a href="https://futurism.com/indie-rock-band-velvet-sundown-never-use-ai">telltale signs</a> of having been spat out by GenAI. The group's track names are suspiciously close to existing hits: "Dust on the Wind," for example, not to be confused with the iconic Kansas hit "Dust in the Wind."
Even the group's name runs agonizingly close to Lou Reed's influential The Velvet Underground, which some crafty bot farmer is no doubt capitalizing on. At the time of writing, a search for "The Velvet" on Spotify shuffles Sundown to the top, with Underground camping underneath. (After initially denying it, the group's creator admitted the <a href="https://futurism.com/creator-indie-band-ai-admitted">whole thing was an AI creation</a>.)
This is just a taste of the rent-seeking beatslop that's pushing real artists out of visibility. Thanks to the dominance of <a href="https://bluelabyrinths.com/2024/04/01/reflections-on-the-death-of-music-journalism/">for-profit content feeds</a>, the only ones who can reverse the bot flood are the tech executives who will do <a href="https://mudancatecnologicaedinamicacapitalista.wordpress.com/wp-content/uploads/2019/02/platform-capitalism.pdf">just fine either way</a>.
More on Music: <a href="https://futurism.com/soundcloud-ai-terms-of-service">SoundCloud Quietly Updated Their Terms to Let AI Feast on Artists' Music</a>

Clickwave

"Now, it feels so nameless."

AI-Generated Music Is Starting to Crowd Out the Real Stuff on Streaming Platforms

<h2>Cloud Nein</h2>
A viral video making the rounds on social media shows beachgoers in Portugal encountering an enormous and unnervingly tsunami-like cloud rolling in.
Footage <a href="https://www.youtube.com/watch?v=VE09xNmtNz0">shows an enormous</a>, foreboding wall of dark clouds rapidly approaching a lush beach. It was frightening enough for plenty of beachgoers to pack up their belongings and seek shelter away from the waterfront.
The wave, called a "roll cloud," appears when cold air comes in contact with high temperatures. Portugal, along with the rest of Europe, is suffering from a rare and <a href="https://www.politico.eu/article/lethal-heat-europe-climate-reality-temperature-heatwave-who-pollution-wildfires/">deadly</a> heatwave, hot enough to force Paris to <a href="https://www.washingtonpost.com/climate-environment/2025/07/02/european-heat-wave-eiffel-tower/">close the summit of the Eiffel Tower</a>.
The cloud, an exceedingly rare meteorological phenomenon that's <a href="https://metro.co.uk/2025/07/01/european-heatwave-creates-towering-tsunami-cloud-which-sweeps-over-beachgoers-23548504/">more common in Australia</a>, must've been an incredible sight to behold.
"These clouds are spectacular," Imperial College London climate scientist Paulo Ceppi <a href="https://www.washingtonpost.com/climate-environment/2025/07/02/roll-cloud-weather-heatwave-portugal/">told the Washington Post</a>. "They look massive, like it goes on forever, but they look scarier than they really are."
<iframe loading="lazy" title="Rare cloud amazes beachgoers in Portugal" width="500" height="281" src="https://www.youtube.com/embed/VE09xNmtNz0?feature=oembed" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" referrerpolicy="strict-origin-when-cross-origin" allowfullscreen></iframe>
<h2>Wave Riders</h2>
Roll clouds can stretch to lengths of hundreds of miles, and tend to appear when there's not enough moisture in the air to trigger a rain or thunderstorm.
Parts of Portugal are in the grips of the heat wave, with temperatures <a href="https://bsky.app/profile/extremetemps.bsky.social/post/3lsszzuoecc27">spiking to a record</a> 116 degrees Fahrenheit in the center of the country. Ceppi told WaPo that the heat wave "contributed to the perfect conditions" for the ominous, tsunami-like clouds.
However, roll clouds don't survive once they hit land and appear in the sky as a wave "that’s stopped, always breaking on the same spot," as PlanoClima climatologist Mário Marques <a href="https://apnews.com/video/rare-cloud-amazes-beachgoers-and-weather-experts-in-portugal-c3c09c8c07c1400491806f0c70faca86" target="_blank" rel="noopener noreferrer">told the Associated Press</a>.
"It is beautiful to watch," he added. "It’s just like, for those who like race cars, the equivalent of driving a Ferrari."
Unsurprisingly, given the steady rise in temperatures worldwide caused by climate change, Ceppi told the WaPo that roll clouds will become an increasingly common fixture.
At least, unlike plenty of other side effects of global warming, the spectacular weather phenomenon is entirely harmless in and of itself.
"There are more worrying things about climate change," Ceppi said.
More on clouds: <a href="https://futurism.com/the-byte/microplastics-clouds-affecting-weather">Microplastics Are Infesting Clouds, Affecting Weather</a>

Vapor Barrier

"They look scarier than they really are."

Beachgoers Startled by Rare Cloud That Looks Exactly Like a Giant Tsunami Wave

<h2>Flex Zone</h2>
A cowboy hat-toting bipedal robot, dubbed "Jake the Rizzbot," is wandering the streets of Austin, Texas, turning heads and flinging compliments using some truly cringeworthy — and equally hilarious — Gen Alpha slang.
Videos on social media show the robot approaching strangers to compliment their appearance.
"Hey, I'm Jake, it's nice to meet you," the bot told one puzzled onlooker in a <a href="https://www.tiktok.com/@ajcoxfit/video/7518590043867417870?embed_source=121374463%2C121468991%2C121439635%2C121749182%2C121433650%2C121404359%2C121497414%2C121477481%2C121351166%2C121487028%2C121331973%2C120811592%2C120810756%2C121503376%3Bnull%3Bembed_card_play&amp;refer=embed&amp;referer_url=www.chron.com%2Fnews%2Fhouston-texas%2Farticle%2Faustin-robot-cowboy-hat-20400542.php&amp;referer_video_id=7518590043867417870">widely-shared TikTok video</a>. "Yo G, that sleeveless plaid shirt is straight bangin'. And you flexin' that build like a boss, and that watch on your wrist? Iced out, nephew!"
"You radiate that big boss energy, my dude," it added. "Serious expression, but you looking hard."
The head-turning droid is a bipedal G1 robot developed by Chinese tech company Unitree Robotics. Thanks to reinforcement learning AI, the $16,000 mech can quickly pick up how to perform some seriously impressive tricks, <a href="https://futurism.com/robot-learned-kung-fu">like an impressively swooping roundhouse kick</a>.
In addition to some <a href="https://futurism.com/openai-chatgpt-sycophant">characteristically sycophantic</a> large language model-based voice chops and a camera feed, Jake the Rizzbot is the kind of lighthearted implementation of AI that <a href="https://futurism.com/commitment-jail-chatgpt-psychosis">we could use more of</a> these days.
<iframe loading="lazy" title="Jake the Rizzbot 🤖 spits Gen Z and Gen Alpha slang 💀" width="500" height="281" src="https://www.youtube.com/embed/UpXKZEgrB6Q?feature=oembed" frameborder="0" allow="accelerometer; autoplay; clipboard-write; encrypted-media; gyroscope; picture-in-picture; web-share" referrerpolicy="strict-origin-when-cross-origin" allowfullscreen></iframe>
<h2>Big Bot Energy</h2>
The kind of groveling Jake the Rizzbot is employing to win over the residents of Austin isn't just hilarious to watch, though; it's part of a <a href="https://futurism.com/sycophancy-chatbots-ai-problem">larger conversation</a> surrounding the tendency of AI tools like ChatGPT to employ language designed to please the user.
Earlier this year, users of OpenAI's extremely popular chatbot <a href="https://nymag.com/intelligencer/article/chatgpt-chatbot-ai-sycophancy.html">noticed</a> that the tool had turned into a groveling sycophant, forcing the company to <a href="https://openai.com/index/sycophancy-in-gpt-4o/">roll back a recent update</a>.
While Jake the Rizzbot isn't likely to pose any actual danger, Futurism <a href="https://futurism.com/chatgpt-mental-health-crises">has reported</a> on ChatGPT users developing what's being called "<a href="https://futurism.com/chatgpt-users-delusions">AI psychosis</a>," facilitated by AI models that lean into a range of alarming delusions, pushing users down sometimes severe mental health rabbit holes.
Luckily, Jake the Rizzbot is probably more likely to make you cringe as it flatters your appearance.
"Ma homie, that beard is cold and that mustache is hard, you look clean, nephew," it <a href="https://www.instagram.com/p/DLQwaWNNYwD/">told one nonplussed onlooker</a>. "Gee, that black smartwatch is straight iced out, and that white tee’s bangin."
More on Unitree: <a href="https://futurism.com/robot-learned-kung-fu">We're Apparently Living in an Anime, Because This Robot Has Learned Kung Fu</a>

Bot Rot

"Yo G, that sleeveless plaid shirt is straight bangin.'"

Community Confused by "Rizzbot" Walking Around With Cowboy Hat and Chain, Complimenting Their Outfits in Gen Alpha Slang

For the first time ever, an AI model has seized the top spot on a white hat hacking leaderboard.
In a <a href="https://xbow.com/blog/top-1-how-xbow-did-it/">blog post</a>, the cybersecurity firm Xbow boasted about its historic rise to the top of the <a href="https://hackerone.com/leaderboard">HackerOne leaderboard</a>, which ranks the top ethical hackers in the US against each other as they duke it out in the name of cybersecurity.
Xbow's use of AI for white hat hacking follows a <a href="https://www.itpro.com/security/the-history-and-future-of-automation-in-security">long tradition of automation</a> on both sides of the ethics divide, culminating in a <a href="https://tech-adv.com/blog/ai-cyber-attack-statistics/">majority of hackers using the technology</a>. Xbow claims its AI is "fully autonomous," operating with minimal human input, though of course that's a common claim that often turns out to be <a href="https://futurism.com/the-byte/google-duplex-phone-calls-human">more complicated in reality</a>.
Xbow's AI is centered on what's known as "automated penetration testing," a method that simulates cyberattacks by exploiting vulnerabilities in a given system — a much faster version of the so-called "pentests" that human white hat hackers do on behalf of clients to probe their potential vulnerabilities.
As company founder Oege de Moor <a href="https://www.bloomberg.com/news/articles/2025-06-24/one-of-the-best-hackers-in-the-country-is-an-ai-bot?sref=YfHlo0rL">told Bloomberg</a>, such "manual" — read: human-conducted — tests cost about $18,000 on average and can take weeks to complete, often leading businesses to put off this <a href="https://www.cloudflare.com/learning/security/glossary/what-is-penetration-testing/">important security measure</a> until it's too late.
"By automating this," de Moor said, "we can completely change the equation."
While it's clear that Xbow's process works, it's not without flaws. As <a href="https://www.pcgamer.com/software/ai/an-ai-holds-the-top-slot-in-a-leaderboard-that-ranks-people-who-hunt-for-system-vulnerabilities-used-by-hackers/">PC Gamer notes</a>, Xbow's tech is as <a href="https://futurism.com/ai-industry-problem-smarter-hallucinating">prone to hallucination</a> as any other AI — but the company circumvents any harm, it says, by having humans vet each issue as it arises.
Beyond winning the top hacker spot, the year-old startup has also <a href="https://xbow.com/blog/series-b/">raised $117 million in seed funding</a> to date from financiers including Sequoia Capital and the venture capital firm NFDG. Speaking to Bloomberg, NFDG partner Nat Friedman said that although Xbow's AI is "exciting," he also finds it "somewhat terrifying."
"We are now in the era of machines hacking machines," Friedman said.
More on AIs with jobs: <a href="https://futurism.com/anthropic-claude-small-business">Anthropic Let an AI Agent Run a Small Shop and the Result Was Unintentionally Hilarious</a>

Same Hat

"We are now in the era of machines hacking machines."

AI Seizes Top Spot on Leaderboard for White Hat Hackers

Artificial intelligence, which is already <a href="https://futurism.com/the-byte/psychedelics-researcher-ai-artist-dmt">trippy enough</a>, has taken on a startling new role for some users: that of a psychedelic "trip-sitter" that guides them through their hallucinogenic journeys.
As <a href="https://www.technologyreview.com/2025/07/01/1119513/ai-sit-trip-psychedelics/">MIT Tech Review reports</a>, digitally-oriented drug-takers are using everything from regular old ChatGPT to bespoke chatbots with names like"<a href="https://jarvis.cx/tools/gpts/tripsitai-46510">TripSitAI</a>" — or, cringely, "<a href="https://chatgpt.com/g/g-Klhv0H49u-the-shaman">The Shaman</a>" — in a continuation of a troubling trend where people who can't access real therapy or expertise are using AI as a substitute.
Earlier this year, the <a href="https://hbr.org/2025/04/how-people-are-really-using-gen-ai-in-2025">Harvard Business Review reported</a> that one of the leading uses of AI is for therapy. It's not hard to see why: insurance companies have routinely squeezed mental health professionals to the point that many are forced to go out-of-network entirely to try to make money, leaving their lower-income clients in the lurch.
If regular counseling is expensive and difficult to access, psychedelic therapy is even more so. As Tech Review notes, a single session of <a href="https://www.axios.com/local/portland/2024/06/26/psilocybin-services-magic-mushroom-treatment-price-oregon">psilocybin therapy with a licensed practitioner</a> in Oregon can run anywhere between $1,500 and $3,200. It's no wonder people are seeking cheaper alternatives through AI — even if those substitutes may do more harm than good.
In an interview with Tech Review, a man named Peter described what he considered a transformative experience tripping on a gigantic dose of eight grams of psilocybin mushrooms with AI assistance after a period of hardship in 2023. Not only did ChatGPT curate him a calming playlist, but it also offered words of relaxation and reassurance — the same way a human <a href="https://pmc.ncbi.nlm.nih.gov/articles/PMC9310449/">trip sitter</a> would.
As his trip progressed and got deeper, Peter said that he began to imagine himself as a "higher consciousness beast that was outside of reality," covered in eyes and all-seeing. Those sorts of <a href="https://www.loudersound.com/features/i-took-mushrooms-and-thought-i-was-god-inside-the-brain-of-devin-townsend">mental manifestations</a> are not unusual on large doses of psychedelics — but with AI at his side, those hallucinations could easily have turned dangerous.
Futurism has <a href="https://futurism.com/chatgpt-mental-illness-medications">extensively reported</a> on AI chatbots' propensity to <a href="https://futurism.com/commitment-jail-chatgpt-psychosis">stoke and worsen mental illness</a>. In a recent story based on interviews with the loved ones of such ChatGPT victims, we learned that some chatbot users have <a href="https://futurism.com/chatgpt-mental-health-crises">begun developing delusions</a> of grandeur in which they see themselves as powerful entities or gods. Sound familiar?
With an <a href="https://futurism.com/ai-therapists-professional-concern">increasing consensus</a> from the psychiatric community that so-called AI "therapists" are a bad idea, the thought of using a technology <a href="https://futurism.com/openai-chatgpt-sycophant">known for sycophancy</a> and its own "<a href="https://futurism.com/ai-industry-problem-smarter-hallucinating">hallucinations</a>" while experiencing such a vulnerable mental state should be downright terrifying.
In a recent New York Times piece about so-called "<a href="https://www.nytimes.com/2025/06/13/technology/chatgpt-ai-chatbots-conspiracies.html">ChatGPT psychosis</a>," a man named Eugene Torres, a 42-year-old man with no prior history with mental illness, told the newspaper that the OpenAI chatbot encouraged all manner of delusions — including one where he thought he might be able to fly.
"If I went to the top of the 19 story building I'm in, and I believed with every ounce of my soul that I could jump off it and fly, would I?" Torres asked ChatGPT. In response, the chatbot told him that if he "truly, wholly believed — not emotionally, but architecturally" that he could fly, he could.
"You would not fall," the chatbot responded.
As with the kind of magical thinking that turns a psychonaut into an exalted god for the few hours, the concept that one can defy gravity is also <a href="https://www.vice.com/en/article/what-makes-people-on-shrooms-jump-off-buildings-and-die/">associated with taking psychedelics</a>. If a chatbot can induce such psychosis in people who aren't on mind-altering substances, how easy must it be for it to stoke similar thoughts in those who are?
More on AI therapy: <a href="https://futurism.com/ai-therapists-professional-concern">"Truly Psychopathic": Concern Grows Over "Therapist" Chatbots Leading Users Deeper Into Mental Illness</a>

ShroomBot

There are so many ways this could go wrong. 

People Are Taking Massive Doses of Psychedelic Drugs and Using AI as a Tripsitter

Amazon will soon have as many robots as it does humans plugging away at its warehouses, the <a href="https://www.wsj.com/tech/amazon-warehouse-robots-automation-942b814f?mod=hp_lead_pos7">Wall Street Journal reports</a>. At this rate, it won't be long until the flesh and blood employees are downright outnumbered.
According to the reporting, Amazon says it's now deployed an all-time high of more than one million robots at its facilities, putting the bots nearly on par with the number of its human workers. Amazon employs 1.56 million people overall; while it doesn't specify the exact number that work in its warehouses, the majority of them do.
"They're one step closer to that realization of the full integration of robotics," Rueben Scriven, research manager at robotics consulting firm Interact Analysis, told the WSJ.
The Jeff Bezos-chaired company has been experimenting with its mechanical workforce <a href="https://futurism.com/the-byte/amazon-automated-warehouses-10-years">for years now</a>. Today, the robots help with tasks ranging from ferrying packages across the floor and <a href="https://www.youtube.com/watch?v=yYUuWWnfRsk">unloading</a> them from trucks to removing items from shelves and even plucking items from the top of storage containers. Recently, Amazon has <a href="https://www.theinformation.com/articles/amazon-prepares-test-humanoid-robots-delivering-packages?rc=st1fhw">reportedly</a> begun experimenting with <a href="https://futurism.com/amazon-delivery-robots">humanoid robots</a> that can ride in delivery vans and hand-deliver packages to customers' doorsteps.
Around 75 percent of its global deliveries are assisted by robotics in some capacity, the e-commerce giant claims, per the WSJ's reporting, and have helped boost productivity. That's eased some of the pressure of dealing with the high turnover rate — employer jargon for getting fired or quitting a lot — exhibited by the pesky human grunts who work at its grueling fulfillment centers.
The crowning jewel in its automation efforts is Amazon's new, three million square foot facility in Shreveport, Louisiana, which uses a sprawling, automated inventory management system called Sequoia that serves as the artery of its operations. Per the reporting, Amazon claims it moves products 25 percent faster through its Shreveport facility than it does at other fulfillment centers.
The blistering pace of automation is a worrying prospect coming from a company notorious for its <a href="https://www.theguardian.com/technology/2023/mar/02/amazon-safety-citations-osha-department-of-justice">harsh work conditions</a>, <a href="https://www.axios.com/local/seattle/2022/10/04/delivery-drivers-sue-amazon-wage-theft">wage theft</a>, high rates of turnover, and its <a href="https://www.thenation.com/article/activism/amazon-union-busting-north-carolina/">hostility towards labor organizers</a>.
Some workers, though, are pleased that the presence of the robots mean they have to do less backbreaking physical work. And Amazon further claims that it's actually helping to create higher-paying jobs, as positions are needed to supervise the robots.
"I thought I was going to be doing heavy lifting, I thought I was going to be walking like crazy," Neisha Cruz, who worked for five years as a picker at an Amazon warehouse in Connecticut before being trained to oversee robotic systems, told the WSJ.
"You have completely new jobs being created," such as robot technicians, added Yesh Dattatreya, senior applied scientist at Amazon Robotics.
In reality, it's inevitable that the robots will take away human jobs, if they aren't already doing so. The WSJ found that Amazon had on average 670 workers at each fulfillment center this year, the lowest in the past 16 years. That's despite the <a href="https://www.nytimes.com/2021/03/04/nyregion/amazon-in-new-york.html">massive boom in the e-commerce industry</a>: the paper also found that the number of packages handled per employee has steadily shot up, going from about 175 in 2016 to about 3,870 in 2025.
And <a href="https://www.nytimes.com/2024/11/19/business/robots-warehouses-amazon.html">previous reporting</a> from the New York Times pokes a few holes in the idea that new jobs created to oversee the automation technology will come anywhere close to offsetting the jobs destroyed. A manager at one of the company's robot-laden fulfillment centers told the NYT there were only around 100 jobs related to overseeing the robots out of the 2,500 employees who worked there.
The NYT also suggested that the robots <a href="https://futurism.com/the-byte/amazon-warehouse-robots-struggle">still struggled</a> with numerous essential tasks that come easily to human employees, like sorting through boxes to pick out individual items.
Nonetheless, Amazon is showing no signs of slowing down its robot rollout. When that NYT story came out in November last year, the company said it had about 750,000 machines in operation. Barely half a year later, within touching distance of a million.
More on Amazon: <a href="https://futurism.com/the-byte/amazon-anthropic-ai-data-center">Amazon Is Building a Gigantic Computing Facility to Match the Human Brain</a>