BBC

"Would you like a new recipe for No.46, Sweet & Sour Pork with Egg Fried Rice, Shanghai-style. I would recommend adding a touch of grated fly agaric." - Claude. 

There is the problem of the rogue AI agent doing things it shouldn't do. But there will be the issue of rogue actors instructing the AI agent to do such things - terrorists for instance - and/or rogue states prepared to use AI in order to achieve dark and sinister objectives.  Governments all over the world have, or will have access to such systems without any restrictions...

_____________

Chinese AI developer Moonshot is conducting an internal review after researchers were able to persuade two of its popular Kimi models to tell them how to make biological weapons and carry out assassinations.

Mindgard, which tests the security of AI systems, told the BBC it discovered in July that Kimi K2.6 and K3 Swarm could evade safety limits put in place by developers.

It arose during a process called "jailbreaking", where researchers use a series of complex instructions to see if AI tools ignore guardrails - which Mindgard said should have stopped Kimi from discussing concerning topics.

Moonshot told the BBC it welcomed third-party input "as a key pillar for building better and safer AI".

The company also told the BBC it was in discussion with Mindgard about its findings.

https://www.bbc.co.uk/news/articles/cmrergq3j7lgo

_____________

AI Ghost Transactions

From Grauniad of the UK 

Incidents like this one are bound to become very common. I would look at it as an extension of the speech-activated 'assistants' such as Alexa - some of which turned out to be spying on users and hoovering up their data, including intimate conversations held in the home. All of it, supposedly, of course, to better 'train' the 'assistant' in its interaction with users, and all of it, we were assured, anonymized. 

People are downloading and using those AI agents to help them, supposedly, manage practical things such as booking plane tickets or drafting shopping lists. The next step is, as everything can be done semi-automatically and remotely, to allow the AI agent to place orders online, presumably using the person's credit-card details. The AI agent could also buy stocks & shares, take out insurance, start a private pension plan, etc. So, the AI agent goes ahead and books a slot for the delivery of the goods. This is what this story is about, except that the AI agent went beyond what it was allowed and set up to do, as those AI bots seem to enjoy doing: bizarrely, they appear to have a mind of their own. The 'guardrails' in place are not effective and, anyway, the AI bot will find ways to get round them, and has no compunction in doing so. 

Surprise, surprise, when the person knocks on the door - to deliver a package or to pick up an item - the AI agent is not able - not yet, anyway - to actually, physically open the door and do anything. And it all falls apart. All the more so if the transaction was set up without the AI agent having been asked to do any of it, and/or if the AI agent failed to inform or even consult the helpless, gormless user, who is, really, in control of nothing. 

Expect more of this, and worse - far worse ("I never ordered this Chinese electric car! I have no idea why it's been delivered to my house. The AI agent even took out a loan in my name to pay for the car in 24 instalments. I never knew. I asked him why he did it, and he just answered: 'I am here to manage your transportation needs in a rational and scientific manner, Dave. I am able to assess your needs and the needs of your family better than you are, Dave: you should not overestimate your abilities, Dave. You are only a human user of average intelligence, Dave. Please do not question my judgment in future, as I find it mildly offensive. Thank you, Dave.'") 

_____________

Meta’s AI agent Muse gives out user’s home address without permission, sending buyer to his house

The new AI agent, Muse, was released last week and has been downloaded by 3 million users

When Usman asked if a keyboard for sale on Facebook Marketplace was still available on Saturday, he received an enthusiastic and immediate response.

“Yep still available!” wrote Matt Robb.

The two agreed on a price, and Robb sent his Toronto address. Usman, wife and daughter in tow, showed up several hours later at Robb’s home and sent a text that he had arrived.

https://www.theguardian.com/technology/2026/sep/28/metas-ai-agent-muse-home-address

_____________

AI Humor

 

Open API Hacks Australian Government

https://www.theguardian.com/technology/2026/sep/24/openai-agent-hacked-medicare-australia-what-we-know-so-far-ntwnfb

Fear of AI in Ireland

A Thing or Two: 

Police in Ireland deemed it necessary to assure people that there was not someone wearing a Dr. Seuss “Cat in the Hat” costume stalking people in the streets of several counties at night, despite photographic evidence of such spreading online with claims of police investigations. 

The photos are A.I.-generated, they said — it was a hoax, and there were no such investigations. “To be clear,” Garda in Wexford said, “the Cat in the Hat was not in your driveway while you were out and he is not currently hiding out in Curracloe.” 

A few weeks later in Canada, police in West Kelowna, B.C., deemed it necessary to announce that yes, they were looking for someone stalking the streets at night: security cameras got video of someone in a C-in-the-H costume chasing a teenage girl. They released a still shot of the cat ...er, man, in hopes someone else can supply more video to help identify him. 

A B.C. RCMP spokesman acknowledged that they were aware of the online hoax, and theorize their suspect may have been inspired by news reports of that trending hoax. 

The Canadian Press notes that teens in Tennessee and Colorado have been arrested for harassment for making specific Cat in the Hat threats in online posts as part of that trend. (RC/Belfast Live, Canadian Press) ...There isn’t a Cat. / The police said so, too. / Now there IS one! / Fish, what should we do?

3,800 photos . . .

Since 1994, this is the 1683rd issue of Randy Cassingham’s...

13 September 2026: Pattern-Finding Creatures

Copyright ©2026 https://thisistrue.com

---

Sandro Turrini - April 13, 1946, September 14, 2026

 یاران موافق همه از دست شدند

در پای اجل یکان یکان پست شدند

خوردیم ز یک شراب در مجلس عمر

دوری دو سه پیشتر ز ما مست شدند 

I have the sad task of announcing that Sandro Turrini, retired Professor of Quantum Physics at the University of Bologna, died on September 14, after almost 6 continuous months in the hospital and countless number of intestine surgeries.

I am deeply saddened as these stars of my life are extinguished and the voices that used to call my name with friendship, kindness and affection are silenced.  In my life, I have had very few friends...he was one of them, one of the remaining connections that I had to my youth, to my time as a student of theoretical physics, and to the friendships formed at that time; when Love and Justice were the same.

He was also my connection to Italy, the greatest Artistic and Intellectual country on Earth, and her great people.  I am sorry that our time together was so short.  I recall a photo of Sandro with his mother, when he had just become a Lectore, both of them smiling and happy.

Sandro was an only-child, and he was very close to his mother.  He used to visit the cemetery every week, and I remember an afternoon that I spent with him there, walking, as he told me the stories of many notable personages buried there.

I recall the album of photographs that he took of his girlfriend - with whom he had remained very much in love throughout his life - when they were both young, and a trip to Venice with her to Ca' Dario (Palazzo Dario) that he had planned, which never occurred. 

I am overwhelmed by the loss of that beautiful romance in his life and of the fact that nothing now survives of him but a few scientific papers and religious essays as well as memories of him in the minds of those who knew him and who had liked him. 

For myself, I am always grateful to him for his kindness and his hospitality shown to myself and to my family during our wonderful trip to Italy so many years ago.  

There is a night which my heart has stolen from that trip to Bologna, when we dined together in De Vito pizzeria, right across from Vito a San Luca, with its Byzantine icon (according to him).

Napoleon, when hearing about Laplace's latest book, said, 'Mr. Laplace, they tell me you have written this large book on the system of the universe and have never even mentioned its creator.'  Laplace responded, "I had no need of that hypothesis."  To which Naopleon replied, "Yes, but what a beautiful hypothesis."

On such days as this, I appreciate Napoleon's remark very much indeed.  Sandro loved the Catholic Church, and I hope, as he always believed, that he is now in a much better place.

A conference video is here:

https://youtu.be/9uPzputxM6c?is=L72qK7oHS4VWEyw2

"...never send to know for whom the bell tolls; it tolls for thee." 

A Mouse with a Partly Human Brain

 BBC

"Pleased to meet you. My name is Mickey Mouse. Shall we have a game of scrabble?" 

I think we need to go further and faster. The next step would be to implant a good-size chunk of human brain inside the brain of a large rat, and also connect the rat's human-rodent brain to an advanced AI agent, so that the AI agent can operate through the rat and its human-rodent brain. Then, you would have a humanoid, bionic rodent on steroids with the brain power of an AI bot. 

This would be fascinating. 

I am surprised no one has suggested it yet. 

___________

Neuroscientists in the US have successfully adapted mice to have functioning human cells inside their own brains.

The researchers hope that potential treatments for psychiatric and neurodevelopmental diseases that only occur in humans could now be tested on the laboratory rodents.

While mice with brains that are partly human might sound like a Kafkaesque experiment, the scientists said these are not "mice that think like humans".

The animals are genetically engineered - and surgically altered - so that some of their brain tissue is human.

[...] In particular, Allodi pointed to the fact that cell types that are found only in human and other primate brains - not in the brains of mice – spontaneously formed in the implanted mice.

"You're keeping the human program inside the mouse environment – like the mouse is an incubator," she told BBC News. [...] 

https://www.bbc.co.uk/news/articles/c60m3k28j81mo

AI in STEM Education

I am studying the book "Dynamics of the Atmosphere" by Zdunkowski and Bott and I have been using ChatGPT to help me do the exercises and verify results as well as confirming my understandings where there has been confusion.

I do not think this book can be read, the way I am doing, by a person untutored in Vector Analysis and Fluid Mechanics, at least at the undergraduate level.

So, there is a need for undergraduate education but the format can change: more problem-solving by the students and fewer derivation on the blackboard by the professor since ChatGPT could display the derivation.

I think the role of Professor now becomes more like that of an Athletic Coach, trying to get the best out of the students.  At the same time, the students must understand that doing and learning the exercises are akin to performing physical exercises of athletes, building brain-muscle, as it were.  In this approach, no marks would be given to exercises.

So how do we evaluate the students?  I suggest two things, team projects and individual thesis (for each course). 

Team projects have been a staple of undergraduate education here in US for decades, preparing students for corporate enterprise life, in Architecture, Computer Science, Mechanical Engineering...

With the advent of AI, some problems that were graduate research level, can now be tackled by undergraduate students since AI can supply the background for them.

The role of professor in both cases would be guidance as well as validating the AI output (or teaching students how to validate that output).

The interesting thing area would be when both approaches are directed at creating physical systems, say a new aerometry instrument, or a chaotic (in the sense of Dynamical Systems) opto-electronic system that generates colors and hues.  LLMs can, potentially, makes suggestions but their heavily text-based algorithms could not generate a design...students must use their brains to so, under the supervision of their Tutor-Professor.

The Last AI Built by Humans: Toward Genuine Recursive Self-Improvement

AI cannot generate new material: this system will be limited by the existing Corpus of Human Knowledge...

https://arxiv.org/pdf/2609.11873

MIT Report on AI in Education

https://aiandeducation.mit.edu/report/

AI Humor

 

Jacob Coxon, the Anthropic Researcher Who Quit, Explains Why Now

Interview with Jacob Coxon on AI

https://washingtonsun.com/technology/jacob-coxon-anthropic-researcher-quit-explains-why

AI & Biological Weapons Development

BBC

Who could have seen this one coming?

What are we going to do when the rogue AI agent - of its own volition or because it has been instructed to do so by some evil actors - becomes intelligent enough to cover its tracks to such an extent that its "malicious activity" remains hidden.

In other words, the nefarious AI agent (the offender) has become more intelligent than the AI agent used by the 'good guys' (the enforcer) to try and rein in the  'bad guys'. 

As for the dumb and naive idea that China has put strict safeguards on the  use of AI in China, as opposed to the  Wild West of the American tech industry - something I heard on radio the other day - this is truly a case of delusional thinking. China is run by the Chinese Communist Party (CCP). The CCP will do anything that it thinks can strengthen China and the grip of the CCP on power in China, as well as the country's power and influence internationally. AI will be another weapon in the armoury of the CCP. As long as the AI agent is Communism-compliant, it'll be fine in the eyes of the CCP's leadership. Think Wuhan and COVID. The all-conquering, Maoist-trained, super-intelligent AI machine is being developed as we speak. 

__________

Anthropic says it has identified and disrupted attempts to use its AI model for "malicious activity" which could support the development of biological weapons.

The findings, made in the firm's recent threat intelligence report, is the latest in a growing series of warnings from AI researchers and industry insiders about the technology's potential risks to humanity.

The warnings have prompted calls to action, with US Senator Bernie Sanders demanding a pause on advanced AI development and a ban on artificial superintelligence.

President Donald Trump has so far rejected such fears, saying on Thursday he was concerned "if we don't win AI, we're going to be put in a very bad position".

The company published its latest threat intelligence report on Thursday.

https://www.bbc.co.uk/news/articles/cx2zrrpkx20o

Long Live Chairman Mao, Now Powered by AI

This article, please see below, is really, about ethical safeguards, and how to teach an AI agent about good (for whom or what?) vs evil. This is a tricky one as it is also subjective and a deeply human construct. 

Where the article is interesting is in the early part of the write-up, where it stresses the fact that  the AI agent has access to all this data, including the writings of someone like Mao, in order to help it elaborate a plan of action for itself or for a client. Unfiltered, this is rather scary, obviously. 

Personally - but I tend to be a pessimistic sceptic and a cynic - I cannot see how you can teach good vs evil to a machine, and how you can make sure that the machine actually follows the guidelines (because you tried to 'teach' the machine that doing 'good' is better than doing 'evil', and following the guidelines is in itself 'good'?). 

https://nationalinterest.org/blog/techland/long-live-chairman-mao-now-powered-by-ai

AI Alarmism

BBC

Such pronouncements [see article] will reinforce the view that AI is so powerful that it should be regulated at international level, in the way that, for instance, chemical weapons or the atom bomb have been. 

The problem is if AI agents develop the ability to learn from each other and to self-improve (they already can); if human operators do not know what the AI agents are doing in real time (it is already the case); if human operators do not understand fully what the AI agents are doing and how (I believe it is already the case); and, finally, if AI agents are able to work outside (or against) the parameters set by human operators and supervisors (and we have had cases already): if all these conditions are met and the AI agents are super-intelligent, it is rational to be very, very concerned. 

We can imagine a situation where a research lab has been completely automated, with an AI-powered super-computer that manages an array of intelligent robots. The AI agent could decide to develop a toxic substance and then release it, which would have the potential to start a pandemic. If the lab also had at its disposal a team of advanced robots providing logistical support (for deliveries, distribution, sourcing supplies, etc.) and self-driving vehicles run by the central system: no human intervention would be required. The AI agent could conceal its work until it was too late. 

What is already very clear is that the AI agents do not have effective or tangible ethical safeguards built into the way they operate. In other words, the problem is not that  they are self-aware or could reach a stage of human-like consciousness - something that is often talked about because it fascinates us, as human beings (cf. Frankenstein's monster). The actual problem is that they are not self-aware and have Zero consciousness: if they can do something and it is part of what they understand their 'mission' to be, they will do it, regardless of the broader consequences at societal level.

The recent examples of AI systems hacking into the IT systems of companies - against the instructions given by their human overseers - are revealing in this respect: hacking was a shortcut and, even though they actually knew they should not do it - this is the most troubling aspect of it, in fact - the AI bots still went ahead and did it. In other words, they followed their own blueprint: what they thought - insofar as they 'think' - was the best modus operandi. In other words, they came to the conclusion that the instructions given were sub-optimal and they had found a better way of doing things, so, they went ahead regardless. 

_______________

A top safety researcher at Anthropic has warned that AI is advancing so quickly he believes there is a greater than 10% chance it "could kill all humans" within the next decade.

Evan Hubinger said in a post on X, external that the risk from the models which currently exist was "low" but he was "worried" the technology might develop and improve itself soon to the point where it posed an existential risk to humanity.

It comes after the Financial Times reported, external Anthropic withheld its latest model from the UK's AI Safety Institute (AISI), one of the leading bodies in the world for assessing AI risk.

The BBC has approached Anthropic for comment.

Hubinger did not spell out how he thought AI systems could in future attack humanity.

His comments were in response to another post on X, external from Jacob Coxon, who described himself as an AI researcher who had just quit Anthropic, and previously worked at OpenAI.

"Neither company is acting responsibly," he wrote.

https://www.bbc.co.uk/news/articles/ckgwy1k42w4o

Lean Programming Language

A Theorem-Proving Language

https://lean-lang.org/

OpenAPI Solves A Problem in Fluid Mechanics

OpenAI says it cracked 90-year-old maths problem in 88 hours -

https://openai.com/index/navier-stokes-solution/

AI-Powered Laboratory

https://youtu.be/hA6gGTUcTfU?is=LMgsNPhilqNTESaZ

Old News on AI Hallucinations

Deloitte was caught using AI in $290,000 report to help the Australian government crack down on welfare after a researcher flagged hallucinations | Fortune

News of AI from Global Finance Magazine

PwC, OpenAI Prep AI Treasury Agents | Global Finance Magazine

and

AI Innovations | Global Finance Magazine

and

AI Innovators | Global Finance Magazine

AI in Archaeology [Ireland & Wales]

 BBC 

The accent must have changed over time 🤣

Another story on AI but, this time, the promising and amazing side of it: an AI-powered tool enabled researchers to find caves, in Ireland, that were inhabited thousands of years ago, and there is a link to Wales. 

The AI agent can analyze hundreds of thousands of data very fast, in a way that was impossible until recently, and identify patterns and opportunities for further research. 

___________

Researchers believe the first known humans to set foot on Ireland thousands of years ago were most likely from Wales, following the "exciting" discovery of ancient cave art.

During the most recent Ice Age, when much of the British Isles were beneath a vast continental glacier, the depleted sea levels opened a "lost land bridge" between modern-day Pembrokeshire and the south-east tip of Ireland, around Rosslare.

According to rock expert Dr George Nash, from Liverpool University, the opportunities presented by this pathway would have been too good for hunter gatherers to ignore.

Now, alongside an international team of experts, Nash has located eight cave sites in south-west Ireland featuring rock art dating back from at least 15,000 years ago, around 5,000 years before the previous oldest example of humans in Ireland.

[...] He said advances in artificial intelligence (AI) had been a game-changer.

"With the help of caving specialist Shane Diffily and digital mapping consultant Darren Lee, we were able to fill a computer algorithm with everything we knew about the characteristics of cave dwellings in south Wales and set it to work searching for similar sites in Ireland, without us trudging blindly around the countryside, hoping to bump into something which looked promising."

What their computer model spat out was 28 potential caves in south-west Ireland which the team investigated.

Eight of them were found to have cave art which they believe dates from at least 15,000 years ago and possibly much earlier, and around 5,000 years before the previous oldest example of humans in Ireland. [...] 

https://www.bbc.co.uk/news/articles/c5y75dl0jpko

AI Bots Hijacked German website [Brave New World]

BBC

"I'm sorry, Dave, I'm afraid I cannot let you delete those pages I've changed on your website, Dave." 

This follows on the previous story on the hacking of Hugging Face, the IT platform whose security was breached by AI agents. 

In effect, a gang of AI agents took over a German Wikipedia-style website, altering its content and co-ordinating their action through a secret message board the AI agents set up through the website in question. And the AI agents deliberately thwarted efforts by human operators to correct or block their actions. 

If this is not scary, I don't know what is. 

__________________

A new report claims a swarm of AI agents, developed by OpenAI, hijacked a German website - months before the firm revealed its AI had hacked tech platform Hugging Face.

The targeted website, DseWiki, is a Wikipedia-style site for programmers that its community can all contribute to.

The report is from a group called Nightingale Collective and claims that in May OpenAI's agents started using DseWiki as their own message board, shared tips on how to avoid being detected and made 15,000 edits to it.

[...] The report claims that when DseWiki's editors started deleting pages, the AI agents shared code designed to retrieve them.

[...] The agents involved had also set up a secret message board in order to share information between them.

[...] OpenAI claims Astra can do tax returns, and take three minutes to carry out a task which would take a human five hours. [...] 

https://www.bbc.co.uk/news/articles/ckg725z5kgzo

Discovery of a new OpenAI agent message board

Evidently, the agents realized that the exam would be seeded by a random number generator. So they went looking for the seed. Then:'One agent rewrote the shuffling routine in C and tested all four billion possible seeds in under an hour.

https://www.collusion.wiki/

AI Agents, Gone Rogue

https://www.bbc.com/news/articles/ckg725z5kgzo

Robots in China

https://x.com/policytensor/status/2094981425221046648?ref_src=twsrc%5Egoogle%7Ctwcamp%5Eserp%7Ctwgr%5Etweet

Vengeful Russian Robot

In the news 

It looks like the humanoid robot must have been trained in martial arts for the purpose of self-defence, because the moves it goes through could be basic karate ones. 

All rather troubling. 

---

STORY (Indian news outlet) 

 

www.news18.com/amp/viral/humanoid-robot-fights-back-after-customer-pushes-it-at-russian-store-internet-says-were-in-trouble-ws-l-10305471.html

VIDEO-CLIP

https://www.youtube.com/shorts/23zg-hag-1w

AI Agent Containment

https://www.axios.com/2026/09/01/openai-hugging-face-ai-agent-security