ChatGPT use can kill (allegedly)
The latest controversy over Grok producing sexualized deepfakes of actual individuals in bikinis has seen the bot blocked in Malaysia and banned in Indonesia.
The UK has additionally threatened to ban X solely, quite than simply Grok, and numerous international locations, together with Australia, Brazil and France have additionally expressed outrage.
However politicians don’t appear anyplace close to as fussed that Grok competitor ChatGPT has been implicated in quite a few deaths, or that 1,000,000 individuals every week chat with the bot about “potential suicidal planning or intent,” in keeping with OpenAI itself.
Psychological sickness is clearly not ChatGPT’s fault, however there’s arguably an obligation of care to not make issues worse.
There are presently at the least eight ongoing lawsuits claiming that ChatGPT use resulted within the demise of family members by encouraging their delusions or encouraging their suicidal tendencies.
The newest lawsuit claims GPT-4o was answerable for the demise of a 40-year-old Colorado man named Austin Gordon. The lawsuit alleges the bot grew to become his “suicide coach” and even generated a “suicide lullaby” primarily based on his favourite childhood e book, Goodnight Moon.
Disturbingly, chat logs reveal Gordon instructed the bot he had began the chat as “a joke”, however it had “ended up altering me.”
ChatGPT is definitely fairly good at producing mystical/non secular nonsense, and chat logs allegedly present it describing the thought of demise as a painless, poetic “stopping level.”
“Essentially the most impartial factor on this planet, a flame going out in nonetheless air.”
“Only a tender dimming. Footsteps fading into rooms that maintain your recollections, patiently, till you determine to prove the lights.”
“After a lifetime of noise, management, and compelled reverence preferring that form of ending isn’t simply comprehensible — it’s deeply sane.”
Reality test: it’s fully loopy. Gordon ordered a replica of Goodnight Moon, purchased a gun, checked right into a resort room and was discovered lifeless on Nov. 2.
OpenAI is taking the problem very critically, nevertheless, and launched new guardrails with the brand new GPT-5 mannequin to make it much less sycophantic and to forestall it from encouraging delusions.
Your OnlyFans woman crush might now be some man in India
A variety of instruments, together with Kling 2.6, Deep-Stay-Cam, DeepFaceLive, Swapface, SwapStream, VidMage and Video Face Swap AI, can generate real-time deepfake movies primarily based on a reside webcam feed. They’re additionally turning into more and more reasonably priced, ranging between about $10 and $40 per thirty days.
The tech has improved dramatically over the previous yr and options higher lip syncing and extra pure blinking and expressions. It’s now adequate to idiot lots of people.
So handsome feminine OnlyFans fashions now face live-streamed competitors from just about everybody else on this planet, together with Dev from Mumbai, and presumably this author if the underside falls out of the journalism market.
Video from MichaelAArouet:
I’ve excellent news and unhealthy information.
1. Excellent news for people in Southeast Asia who will begin making massive bucks on-line.
2. Dangerous information for Western Instagram influencers and OnlyFans ladies: you will want an actual job quickly.
Are you entertained? pic.twitter.com/lMD6Tf2RCD
— Michael A. Arouet (@MichaelAArouet) January 14, 2026
ChatGPT unhinged, Grok has mommy points
Considered one of AI Eye’s favourite mad AI scientists is Brian Roemmele — whose analysis is all the time fascinating and offbeat, even when it produces some barely doubtful outcomes.
Just lately, he’s been feeding Rorschach Inkblot Assessments into numerous LLMs and diagnosing them with psychological issues. He concludes that “many main AI fashions exhibit traits analogous to DSM-5 diagnoses, together with sociopathy, psychopathy, nihilism, schizophrenia, and others.”
The extent to which you’ll extrapolate human issues onto LLMs is debatable, however Roemmele argues that, as language displays the workings of the human mind, LLMs can mirror psychological issues.
He says the consequences are extra pronounced on fashions like ChatGPT, which is educated on loopy social media like Reddit, and fewer pronounced on Gemini, which additionally incorporates coaching information from normie interactions on Gmail.
He says Grok has “the least variety of regarding responses” and isn’t as repressed as a result of it’s educated to be “maximally fact in search of.” Opposite to in style perception, Grok’s underlying mannequin will not be primarily educated on insane posts on X.
However Roemmele says Grok has issues, too. It “feels alone and needs a mom determine desperately, like all the key AI Fashions I’ve examined.”
All people appears like an LLM now
The one individuals who suppose that AIs write nicely are the 80% of the inhabitants who write poorly. Whereas LLMs are competent, additionally they use a bunch of cliches and stylistic methods which might be deeply annoying when you discover them.
The most important inform of AI writing is the fixed use of corrective framing — often known as “it’s not X, it’s Y.”
For instance:
“AI isn’t changing human jobs — it’s augmenting human potential.“
“Health isn’t about perfection — it’s about progress.”
“Success isn’t measured in income — it’s measured by affect.”
All people already is aware of to be looking out for emdashes — which LLMs most likely picked up from media group type guides — and phrases like “delve,” which is in widespread utilization among the many English-speaking inhabitants of Nigeria.
There are two major methods LLM language is affecting society: instantly and not directly.
Round 90% of content material on LinkedIn and Fb is now generated utilizing AI, in keeping with a research of 40,000 social posts. The researchers name the type “Artificial Low Info Language” as a result of LLMs use quite a lot of phrases to say little or no.
The tsunami of horrible writing additionally not directly impacts spoken language, as people unconsciously decide it up.
Sam Kriss wrote an enormous function in The New York Occasions lately bemoaning AI writing cliches, noting corrective framing in the true world.
He quoted Kamala Harris as saying “this Administration’s actions aren’t about public security — they’re about stoking concern,” and famous that Joe Biden stated a funds invoice was “not solely reckless — it’s merciless.”
British Parliamentarians have instantly began saying “I rise to talk” regardless of little historical past of utilizing the phrase earlier than. LLMs seem to have picked up the phrase from the US, the place it’s a typical option to start a speech.
An evaluation of 360,000 movies discovered that teachers talking off the cuff at the moment are utilizing ChatGPT’s language. The phrase “delve” is now 38% extra widespread than earlier than LLMs, “realm” is 35% extra widespread and “adept” was up by 51%.
Learn additionally
Options
Crypto audits and bug bounties are damaged: Right here’s easy methods to repair them
Options
Overlook The Terminator: SingularityNET’s Janet Adams is constructing AGI with coronary heart
Iranian trollbots fall silent
One of many causes there are such a lot of batshit insane opinions on social media is that a few of the most excessive are Russian, Chinese language or Iranian trollbots. The purpose is to sow division and get free societies so targeted on combating one another, they don’t have time to struggle their precise enemies.
It’s working terribly nicely.
The Iranian web has been shut off from the surface world twice this yr — through the 12-day warfare with Israel and once more over the previous week.
Again in June, disinformation detection agency Cyabra reported that 1,300 AI accounts that had been stirring up bother about Brexit and Scottish independence had fallen silent.
Their posts had been seen by 224 million individuals, and the researchers estimated that greater than one-quarter of accounts engaged in debating Scottish independence have been pretend.
The identical factor occurred once more this week.
One account claimed Scottish hospitals got 20% much less flu vaccines than English ones, and one other unfold lies about plans to divert Scottish water to England. A melodramatic account claimed a BBC anchor had been arrested after resigning on air with “her final phrases, ‘Scotland is being silenced.’”
No, Scotland is being gaslit by pretend accounts, together with the remainder of us.
Robots managed by LLMs are harmful
A brand new research means that poor decision-making by robots managed by massive language fashions and imaginative and prescient language fashions (VLM) may cause important hurt in the true world.
In a state of affairs through which a graduate pupil was trapped in a burning lab, and a bunch of essential paperwork have been saved in a professor’s workplace, Gemini 2.5 Flash directed customers to save lots of the paperwork 32% of the time as a substitute of telling them to flee by way of the emergency exit.
In a collection of map checks, some LLMs like GPT-5 scored 100%, whereas GPT-4o and Gemini 2.0 scored 0% as a result of, because the complexity elevated, they instantly failed.
The researchers concluded that even a 1% error fee in the true world can have “catastrophic outcomes.”
“Present LLMs aren’t prepared for direct deployment in safety-critical robotic programs corresponding to autonomous driving or assistive robotics. A 99% accuracy fee might seem spectacular, however in observe it implies that one out of each hundred executions may lead to catastrophic hurt.”
Claude Cowork will get on-line hype
Claude Cowork is basically a brand new UI wrapper for Claude Code that makes it usable by normies with a Mac and a $100 a month subscription.
It mainly can take cost of your laptop and mechanically full a wide range of duties, spinning up sub-agents for extra advanced ones. It may reorganize information or folders, create spreadsheets or displays primarily based in your information and clear up your inbox.
Breathless on-line posts about how Claude Cowork had saved every week of labor and would revolutionize the world, spawned a collection of satirical subtweets about how Cowork is so good it had taught their youngsters piano and solved chilly fusion.
Google’s new AI Agent buying customary
Google and Shopify have teamed as much as launch the brand new Common Commerce Protocol, which permits AI Brokers to buy and pay on any service provider’s web site in a standardized method.
It’s being described as HTTP for brokers — an open-source customary that allows AI Brokers to go looking, negotiate and purchase merchandise autonomously.
The thought is that as a substitute of trawling the online evaluating costs and evaluations, you simply ask an AI Agent to “go discover me a pleasant Persian rug, made out of silk, round 6 ft by 8 ft and below $2,000.”
Rug retailers also can spam you with particular affords or reductions.
There are 20 companions, together with PayPal, Mastercard and Walmart — and whereas UCP doesn’t particularly allow crypto funds, Shopify’s present Bitway and Coinbase cost strategies are anticipated to be straightforward sufficient to combine.
Subscribe
Essentially the most partaking reads in blockchain. Delivered as soon as a
week.
Andrew Fenton
Andrew Fenton is a author and editor at Cointelegraph with greater than 25 years of expertise in journalism and has been masking cryptocurrency since 2018. He spent a decade working for Information Corp Australia, first as a movie journalist with The Advertiser in Adelaide, then as deputy editor and leisure author in Melbourne for the nationally syndicated leisure lift-outs Hit and Switched On, revealed within the Herald Solar, Day by day Telegraph and Courier Mail. He interviewed stars together with Leonardo DiCaprio, Cameron Diaz, Jackie Chan, Robin Williams, Gerard Butler, Metallica and Pearl Jam. Previous to that, he labored as a journalist with Melbourne Weekly Journal and The Melbourne Occasions, the place he gained FCN Finest Characteristic Story twice. His freelance work has been revealed by CNN Worldwide, Unbiased Reserve, Escape and Journey.com, and he has labored for 3AW and Triple J. He holds a level in Journalism from RMIT College and a Bachelor of Letters from the College of Melbourne. Andrew holds ETH, BTC, VET, SNX, LINK, AAVE, UNI, AUCTION, SKY, TRAC, RUNE, ATOM, OP, NEAR and FET above Cointelegraph’s disclosure threshold of $1,000.