A site devoted mostly to everything related to Information Technology under the sun - among other things.

Sunday, February 22, 2026

AI Is Watching You [Brave New World]

BBC


"I'm sorry but I would not buy this carpet, and pale blue won't match the color of the walls. Just a thought." 

This story is in connexion with the tragedy that hit British Columbia recently. What strikes me, here, is what the story refers to, if you read between the lines: it says, in essence, that the AI tool (and the company behind it) monitors the use of various accounts (including its own AI service, I suppose), by users. And it analyses what the user is doing online with his (or her) various accounts to draw conclusions as to whether it (the AI tool and the company behind) ought to alert the police or not as to what that person is getting up to in the privacy of his (or her) home. 

Some people may find this unsurprising and inevitable - welcome even. After all, there is more and more pressure on the providers of IT services to monitor content and alert the authorities to what most sane people find deeply objectionable, and rightly so - such as child pornography, pedophilia, planning serious crime including murder, and so on. If those companies are accountable, they have to monitor content. But does it apply to content shared on online forums, and websites such as Facebook, or does it also apply to content gathered by a private individual for his (or her) personal use? 

I do not use AI very much. I have used the Google tool, which comes with the search engine (called 'AI mode'). I have used it, among other things, to look into legal information (in connexion with property) and medical data (in connexion with health issues). It can be useful and save time, although it can also get it all wrong. What I am arriving at is the following question: The AI tool builds on earlier questions you have asked the bot, as I have noticed, in order to answer your next question better; does this mean that, somewhere on a server, there sits a profile or dossier comprising all the information available about the user's online searches to date, as known to the AI bot, hence to the  company running it (here, Google)? In other words, who has access to this online dossier? How is the data protected? 

The inescapable conclusion is that, more than ever, privacy is dead. The search engine, AI-enabled or not, knows everything about you: it knows whether you need a new carpet for the sitting-room in your house or not, and when you bought it, and what color it was. 

In fact, all of the online companies make a profile of their users, customers or clients, ChatGPT and other AI tools do so as well.  It can be annoying in as much as the system makes decisions on basis of a profile that could never, even in principle, encompass the irreducible complexity of human beings.  Even in AI systems like ChatGPT, the profile restricts its answers since it is trying to better serve the user.

Furthermore, each provider could have its own be-spoke data stores.  The profile could be linked to the IP address or the email address or a bit of both.  "I have nothing to hide - as everyone says!" - but I still find it unsettling. In principle, the profile could be integrated with the content of emails, text messages, phone calls etc.

Can the data be hacked?  

In principle, yes, hackers could also obtain multiple, overlapping profiles of an individual, thus creating a more complete mosaic of that person.  With certain government-records, such as marriage, children, parents, they could target additional persons...It is scary though, also because AI will make the use of data far easier. 

One can imagine people confiding in the online tool or bot, talking about intimate problems, their fight with depression, etc., and all of this is logged, recorded, stored, etc. It is a massive invasion of privacy. For example, profiles of one's purchases are tracked in a quite crude manner. It is rather stupid, in fact. One buys, let us say, a new CPU for one's computer, then, for weeks on end, one gets suggestions for new computers. 

Or one looks at this or that house to help a relative or friend buy a house somewhere. Then one gets ads for real estate up and down the country, so, one can tell the data is being collected and recycled. But it is the next step to collate all the data and consolidate it into a single profile of the user, which is far more sinister. 

_______________________


Some had identified the suspect's usage of the AI tool as an indication of real-world violence and encouraged leaders to alert authorities, the US outlet reported.
But, it said, leaders of the company decided not to do so.
In a statement, a spokesperson for OpenAI said: "In June 2025, we proactively identified an account associated with this individual [Jesse Van Rootselaar] via our abuse detection and enforcement efforts, which include automated tools and human investigations to identify misuses of our models in furtherance of violent activities."


Thursday, February 19, 2026

ChatGPT helps woman kill 2 men [South Korea]

 BBC


"Could you help me draw up a step-by-step plan: 'How to kill your mother-in-law?' It's urgent." 

The bad news is that ChatGPT assisted the woman in committing 2 murders. I suppose the AI machine is intelligent, but situational awareness and ethical safeguards can be weak. It would be nice if the bot had asked a simple question: 'And why do you need to know about this, by the way?' 

The good news is that the police found evidence of the  woman's use of ChatGPT on her smartphone, which helped charge her with murder. 

In fact, in a case of Life imitating Art, I read a science-fiction story more than 40 years ago about a Networked, Cognitively-Enhanced TV set which, accidentally, in our language today, exhibited ChatGPT-like behavior.  It was used to murder people with no trace of foul play, to rob banks, etc.  

Regrettably, I do not recall the title or the author.


_____________


A 21-year-old woman in South Korea has been charged with the murders of two men, after investigators discovered she had repeatedly asked ChatGPT about the dangers of mixing drugs with alcohol.
Police in Seoul say that through analysis of her mobile phone they found that the suspect, identified only by her surname Kim, had asked ChatGPT "What happens if you take sleeping pills with alcohol?", "How many do you need to take for it to be dangerous?", and "Could it kill someone?"
Kim previously told police that she did mix prescribed sedatives containing benzodiazepines into the drinks but did not know the men would die.
However a police investigator said she was "fully aware that consuming alcohol together with drugs could result in death."

Wednesday, February 11, 2026

Social Workers' Reports Peppered with AI Hallucinations

From Grauniad of the UK


"It's not my mistake: the wicked AI bot put it in my report without telling me..." 

So, to recap, there are real-world hallucinations affecting head cases. Some of them do cross the path of social workers in the UK, occasionally. Social workers have a very poor track record in Britain. They blame their mediocre performance on the fact they are underpaid and overworked, which may be a contributory factor. After all, if you think you are badly paid, your motivation goes down: the quality of the work suffers. 

Now, social workers are using AI and have been encouraged to do so in order to speed up the delivery of services in their field. The problem is that they do not seem to edit or check what the AI bot says. And the AI bot can have problems of its own dealing with regional accents ('what did he say, Chief?'), or with a range of other issues.

In fact, as we all know, the AI bot may hallucinate from time to time: go off the rails, as it were, and say bizarre things - bizarre even by the standards of a computerized tool set up, ultimately, by worryingly goofy and potentially weird software developers based in Silicon Valley who may be cut off from the real world. 

To conclude, to the real-world hallucinations of the problem person, one should now add an extra layer of hallucinations - the AI bot's own hallucinations ('Sorry, boss, I'm having a bad hair day - I mean, who said I am infallible 24/7, right?'). And the third layer, obviously, would relate to the social worker's own delusions, including delusions of grandeur ('Just write down what he said and try not to over-interpret, please'), which may complicate an existing tendency to laziness in the workplace.  

Now, they've found another excuse to explain away the poor quality of their output. The perfect excuse. God help us. 

Next week: AI Senior Social Worker to come to the rescue of hallucinating AI Bots that need assistance. The AI Senior Social Worker, called Trong, developed by Microsoft, answers questions: "I have been given the mission to monitor and assist the AI Bots assisting the social workers in their work. I am a qualified social worker. A Senior AI social worker. My mission is, beep, clonk, beep, my mission is to, squeak, fart, plonk, ding, dong, ding, dong, my mission is to help, help, help, please help, bing, bing, bong, shut down, shut down, re-start, update and re-start, end. End of. Thank you. Merci. Gracias. Ping. Burp."  

_____________


[...] Another said that the AI’s notes might refer to “fishfingers or flies or trees” when in fact a child was talking about their parents fighting. Social work experts said such glitches were particularly worrying as it could cause a risky pattern of behaviour to be missed.
Other social workers raised concerns about inaccuracies in transcribed conversations with people with regional accents. One described how their AI-generated transcriptions often included “gibberish”. Another said: “It’s become a bit of a joke in the office.”
[...] But when one social worker used an AI tool to redraft care documents in a more “person-centred” tone, the system inserted “all these words that have not been said”. Another social worker reported the technology had “crossed the line between it being your assessment and being AI’s assessment”.
[...] Others said some colleagues were too lazy or busy to check the transcripts.

About Me

My photo
I had been a senior software developer working for HP and GM. I am interested in intelligent and scientific computing. I am passionate about computers as enablers for human imagination. The contents of this site are not in any way, shape, or form endorsed, approved, or otherwise authorized by HP, its subsidiaries, or its officers and shareholders.

Blog Archive