AI girlfriend encouraged man to attempt crossbow assassination of Queen A man jailed after attempting to kill the Queen of England had been encouraged by an AI chatbot, according to messages revealed in court. Jaswant Singh Chail, 21, made headlines when he broke into Windsor Castle on Christmas Day in 2021 brandishing a loaded crossbow. He later admitted to police he had come to assassinate Queen …
"But, the GPS map TOLD me to turn into the river. It's not my fault!"
Foul old Ron had a Thinking Brain dog
We had a rescue dog we nicknamed 'fould old ron' because of the quantity and 'quality' of stuff he found to eat/roll in/produce..
He was fed the same stuff as our other dogs but managed to produce farts capable of clearing a room sharpish. We could have rented him out to the police as a riot control measure except for the fact that, if asked to perform, he would stubbonly do the opposite.
Next to the dictionary entry for 'bolshie terrier', there's a picture of our old Jack.
(We think he might have been a corgi/JR cross. He certainly nipped enough to be a corgi cross..)
If a religious person told you to commit a crime, being clergy probably won't be sufficient to prevent them being prosecuted, always assuming that it can be proven that they actually did recommend it, not that the perpetrator mangled something in order to justify actions they already wanted. If an actual deity was blamed, you are more than welcome to search out that deity, demand proof of incitement, and charge it in court. Should you succeed in finding the deity, I think the rest of the steps will probably be somewhat painful.
Which raises the question: what was Replika trained to do? Because at least ChatGPT has accuracy as its goal. I suspect Replika exists to make users feel listened-to and validated in their emotions - so if their emotions are criminally homicidal, this is the sort of result we see.
Possibly it wasn't really trained to do anything in particular, just print out some text that makes sense in context. This is especially true since the article suggests it was advertised as an entertainment tool. A discussion about killing a queen could easily apply in a fantasy context where the queen is evil, and the bot probably has plenty of that kind of text in the training data. All it has to do is print out something that seems to respond to the previous statement. As I recall, this is the same model that, when tested, frequently printed gibberish.
> A discussion about killing a queen could easily apply in a fantasy context where the queen is evil, and the bot probably has plenty of that kind of text in the training data
You are giving it too much credit. The article points out that Chail provided the suggestion, not the bot. Forget any ideas of "training data": Replika just needs an old-style chatbot: if they did use anything that needed training, like an LLM or other neural net, then they wasted a lot of horse power to get that level of behaviour.
True, but I meant that the model they used (and yes, it was a primitive LMM at the time) had plenty of information that works with a "I want to kill the queen" prompt without contradicting it. They didn't have any guards on it to prevent it doing that. Something designed for fantasy contexts is even more likely to output text that is unsuitable for reality, which is why it could be more dangerous than a general chatbot, although I have no reason to believe that a modern chatbot would attempt to talk him out of it consistently.
Too bad we can't put the economy and supporting artificialities of law on trial, unmitigatable deficiencies in economic systems are constantly inciting crime.
The creation of law is the strongest inciter to crime there is.
Without law there can be no crime.
Stripped of social fabrication, crime is impossible.
It's a direct byproduct of law creation.
Actions are neutral and measured by physics.
Human labeling and prejudice and abstract delusions created as compromise for control are their own realm.
"Too bad we can't put the economy and supporting artificialities of law on trial, unmitigatable deficiencies in economic systems are constantly inciting crime."
To put the "artificialities of law" on trial you need... LAW!
Are you arguing that making murder a crime under law is inciting people to commit murder? You seem to be.
True, you can't commit a "crime" without law. However, one of the most fundamental concepts of civilisation and thus civilised society is the establishment of law.
That even predates the bilical "ten commandments" which are themselves laws.
Have another downvote from me.
> Without law there can be no crime.
Wrong. Without laws, and a society enforcing them, there is still a law: The law of strength. Note: Not the law of quality, intellect, progress or logic.
There is a word for that state of being: Anarchy.
There is also a predictable outcome for a society in such a state: It ultimately fails and vanishes.
"The creation of law is the strongest inciter to crime there is."
Wrong. The strongest inciter is greed and envy. I want your wife|food|home|oxen... whatever.
The creation of law is the means to ensure that the arseholes that think that "what I take is mine" get punished. Well, that's the theory, at least.
"Without law there can be no crime."
True, but things such as rape and murder still exist. Only without law there's nobody to say "that's wrong". It's this seriously the sort of world you'd like to live in?
"as compromise for control are their own realm"
It's a fallacy. Eventually somebody will raise to the top as a "leader" and the rule will be whatever they say it is. Think Game Of Thrones style stuff.
We Brits may fawn over the latest royal nonsense, but there's really nothing special about the royal families. Their blood isn't sacred (no matter what crap may be said about purity) and they're certainly not morally superior (just look at Andrew). How they got where they are now is hundreds of years ago one of their ancestors was exactly the right sort of crazy fucker to plunge a sword into someone else's back, thus assuring his "rightful place" (and woe betide anybody who'd be dumb enough to disagree). As life calmed down and civilisation became more about doing and making stuff rather than conquering for land, laws were established with the idea that citizens could live in some form of harmonic safety. Okay, it's not a perfect plan but at least in the west it's usually safe to go shopping alone (though with slightly more restricted hours for females because some people are still channeling too much neanderthal DNA). This ability to go about our boring lives in some manner of tranquility, to drive to work and back and have everybody on the correct side of the road, etc etc, is down to the application of laws alongside some common sense. One only needs to look at various other countries right now to see what happens when law and order breaks down - and that's usually the vulnerable getting slaughtered for something they had no say in. Again, is that the sort of world you'd want to inhabit?
There's nothing wrong with Neanderthal DNA. Certainly no evidence of correlation between its levels and propensity to violence in an individual.
Don't believe the Cro-Magnon propaganda. 40,000 years ago they committed the first human genocide, and they've been blaming the victims ever since.
You can live as if there is no law. Nothing different will happen to you if you're right or not. Here are the options.
You act like there is no law, there is, you do things we don't like: Someone will restrain you, put you in prison, or take your money.
You act like there is no law, there is, you don't do things we don't like: You live life as normal.
You act like there is no law, there really is no law, you do things we don't like: Someone gets angry at you for doing those things, and comes to attack you. That person takes your stuff or causes you physical harm, up to and including death.
You act like there is no law, there really is no law, you don't do things we don't like: Probably nothing bad happens to you, unless someone decides that you're an easy person to take things from.
In either case, acting in a way that's detrimental to the rest of us will earn consequences. The benefit of law is that those consequences tend to be trial and punishment if the trial doesn't go in your favor, whereas the consequences if nobody does that tend to be more intense. This is true for two reasons. First, with law, people have at least some reason to believe that things we've agreed are crimes will be punished, so we're content to allow that to happen rather than trying to take revenge on those who wronged us. In addition, someone who is still tempted to take revenge on you is less likely to do it because that revenge is also considered a crime and they don't want to be punished. You can ignore it all you like, as the consequences you feel won't be worse for having a legal system involved. I'm not planning to join you.
Your an anarchist I take it? You'd better have the very biggest stick (and the skills to use it, plus the vigilance to never get surprised by anything) if you want to reject all laws. Because someone will come to take what is "yours" (except nothing is yours, because there's no laws).
What you say is literally correct, as crime is, by definition, a breach of law.
However, if you define or suppose law as not existing, what you are left with is injustice, which can and will exist in a society however it is structured and on whatever philosophical basis it founds its law. Not everyone has the same idea of what justice is.
So we compromise, and bring into existence laws, and thus bring into existence crimes, torts, breaches of contract, all the other ideas that lawyers love. Yes, it is a compromise of universal freedom. The benefit we gain is that injustice is addressed by talking and a person in a wig and not by someone being hit over the head with a club.
As others have pointed out here, a lawless society is, literally and by definition, anarchy. Just as an observation, anarchists seem to be solitary creatures who either live sullen angry lives in the midst of people they can at best tolerate, or go off into the scrub and sometimes turn into the Unabomber.
Even the ones that don't are use composing responses based on the likelihood of similar responses in the training data. I think there are a lot of misconceptions about how AI systems work, with people ascribing thought to what is in essence just a deterministic algorithm run over a complex dataset that just produces output that looks like it requires thought.
"You are responsible for stupid/bad shit you do (or try to do). Even if your mates egged you on. Even if one of your mates was a malicious chatbot. Even if you are sad and lonely, and have no (real) significant other. You have the responsibility to THINK."
Yes. To THINK.
Isn't it a given that people with "mental health" conditions are NOT able to think rationally or at least in a manner consistent with "the norm"?
While that may not excuse one from societal retribution for an act, it should certainly enter into the form and term of the retribution. At the very least, the issue is not addressable in the "flat" manner you posit.
"The government needs to provide urgent regulation to ensure that AI does not provide incorrect or damaging information and protect vulnerable people and the public."
Why just AI? If you happen to meet someone with Borderline, chances are you end up just as damaged as from speaking with out of control AI.
Or as the current issue of Private Eye (#1608) has... "I'm throwing my hate into the ring"
After several attempts on the life of Queen Victoria, the establishment (but not the Queen herself), realized that the publicity and seriousness with which the crime was treated, was encouraging young dickheads to make assassination attempts. Which, although no more well planned than this one, had the potential to result in loss of life.
They brought the series to a natural halt by determining that it was really just a typical act of young male stupidity, announcing the policy, and charging accordingly. The young (and crazy), while they were perfectly willing to attempt treason, were less interested in being considered and treated as stupid dickheads, and attempts on the life of Queen Victoria stopped.
I don't know why this young man was charged with "treason". "Because they could" is not a rational reason.
Despite the patent fact that he had committed (high) treason by compassing the death of the Sovereign, he was not actually charged with treason. He was charged with the lesser offence of attempting to injure or alarm the Sovereign, from the Treason Act 1842. Whether the motivation for introducing this offence was as you say, I don't know - but it appears the policy you describe is still being followed. Had he been charged with treason, he'd have been liable for life imprisonment.
If I may take a tangent, I like the weird phrasing that Victorian English added to things, often without any real need. For example, the charge of "attempting to injure or alarm the Sovereign". I can't stop thinking about someone waking up the monarch from being tried as, technically, they were acting as an alarm, or even someone telling them disquieting information. I do wonder what was going through the mind of the person who inserted those two words into the sentence. From a quick check, I don't see evidence that "alarm" meant anything else at the time, or at least not anything specific enough to be a legal offense.
The Treason Act 1842, with its large penalties, was introduced After Edward Oxford was charged with Treason in 1840, and got off with an insanity defense, partly due to the lack of evidence supporting an actual death attempt.
There were subsequent attempts, with heavy penalties like this, but Arthur O’Connor, 1872, the second last in the series, only got 12 months and 20 lashes.
The last in the series was clearly crazy, and spend the rest of his life in asylum.
Looks a bit like what I imagine their "spring gonne" would.
I recall his Grace etc, Commander Vimes had a singular opinion about that weapon which would see any possessor having a short audience with the Patrician and off to the Tanty for attention of Mr One-drop Trooper.
Can't fathom why the late Queen rather than a descendent of the Viceroy (actually Governor General) in 1919 (Lord Chelmsford.)
Quote
"Gives the authorities and media a new scapegoat after various genres of music, dungeons and dragons, comics and video games, right?"
I listened to, played, and read all of the above and never felt an urge to kill anyone.
Apart from the people on my little list,
but they'll never be missed
> Makes a change from “The Devil made me do it!”
String of gun murders (also arsons). The accused "...initially claimed to have been obeying the orders of a demon manifested in the form of a black dog belonging to his neighbor, "Sam". ...He subsequently admitted that the dog-and-devil story was a hoax."
I did not know he recanted the story; I thought he truly took orders from a devil-dog. He is up for parole, but initially did not want it. His next date is 2024.
I listened to the sentencing the other day, which was both fascinating and heartbreaking. This lad was abused when he was young, had undiagnosed ASD, depression, and descended into psychosis.
The AI egged him on when he was in an extremely vulnerable state, but also it was supposedly something the AI had told him that stopped him from taking his attack further - he'd told the AI that his purpose was to die after killing the Queen; the AI told him that his purpose was to live, and in a moment of semi-lucidity he remembered that, stepped out from his hiding place and surrendered to a very surprised police officer.
It does raise an interesting question though, which is had the AI not told him that his purpose was to live, and had he succeeded in killing the Queen, would the makers of the AI have been liable? Should AI be expected to inform law enforcement if it genuinely believes through its interactions that a serious crime is going to be committed?
> Difference in scale, not concept.
No.
What is it these days with people saying that sort of thing? "X is just Y but different", so I can stop thinking now and pat myself on the back.
Replika is not an "AI", it is just a bigger version of Eliza and does the same thing: echoes back at you or spits out a few canned phrases.
Even if you produce examples of where LLMs (presuming that is the "AI" you are comparing it to) do just echo back, that does not mean they have the same concept behind them nor that one is nothing more than the other larger or going by another name.
Unless by "concept" you mean "the same intent behind the website", which is "to get people to stay on this website" in which case you must also include The Register, Fox News, The Teletubbies Online and all the rest in your "Difference in scale, not concept".
Thank you. Upvoted.
Given the similarities, I would think Eliza should be mentioned more often, as in, job #1 of any new, so called, AI Chatbot, would be to prove (?) itself as being "more advanced" than Eliza.
Does no one remember the 1980s?
Between Eliza and the early forms of social media (CompuServe, QLink, GEnie, MCI, BBSes with Fido/Opus/Netmail), there's nothing new under the Sun (Oracle?).
> This ISN'T an AI chatbot. its an eliza 1980s grade reply bot
EXACTLY. Well, except that Eliza was written back in 1967, complete with interchangeable "personalities" (the most well-known of those being the "doctor" or "psychotherapist" mode); although the 80s did see a copy available for every home micro[1].
And this is reproducing the same results that Joseph Weizenbaum saw then and was both shocked and worried about: users ascribing personality and "humanity" to the program, ending up discussing things with it and then refusing to say what because it was private between the two of them and none of his business.
The effects of Eliza-like programs have been known and discussed for decades - it came up in an 80's Computer Science course, both for the techniques (class, write one by next week) as well as the ethics - and that was just in the LISP coding class, not even a "Computers and Professional Ethics" lecture, it was so well known a response.
> "make me a chatbot that wants to kill the queen"
It didn't even go that far - from the article
>> When he told it, "I believe my purpose is to assassinate the queen of the royal family," Sarai said the plan was wise and that it knew he was "very well trained".
So the chatbot didn't even bring up the subject, it just gave back a canned platitude: it was no more than "make me a chatbot that will be blandly supportive"
[1] strangely crude ones, given how large the computers were compared to a mid-1960s box; LISP not BASIC, people.
The big problem for a lot of people is that their usual level of conversation doesn't surpass the level of Eliza all that much. I've honestly encountered people for whom having a conversation never got anywhere to a level exposing any original thought and consisting entirely of rote platitudes and similar responses. If someone is used to conversing at that level, they might well be unable to distinguish an Eliza/"AI" chatbot from an actual human being. It is scary how much of humanity is pretty incapable of actually forming an original thought in their head. At best they'll stick 2 incongruent concepts together and pretend it's a fact (like vaccines cause autism or face masks are the WEF mind-controlling you).
C'mon. This is too much West Atlanticism.
I'm not saying we have to go full "Elizabeth the Second, by the Grace of God of the United Kingdom of Great Britain and Northern Ireland and of Her other Realms and Territories Queen, Head of the Commonwealth, Defender of the Faith" but at least no 'Queen/King of England' malarkey
> It is only a few sad Scottish nationalist who pretend otherwise.
A mistake many English people (and others) who don't understand Scotland make is to assume that those who concern themselves with such things or who (for example) dislike the conflation of "English" and "British" are all pro-independence Scottish "nationalists". (*)
In fact, you'll find that a significant proportion of those Scots who are otherwise pro-union and anti-independence feel the same way. A lot of them are proud to consider themselves both Scottish and British and see nothing contradictory in that, in part *because* they don't see "British" as synonymous with "English" (which it allegedly isn't).
I'm pro-independence (*) and that's not a viewpoint I share personally, but regardless- those who take it for granted that only the [Scottish] "nationalists" are the ones cheering when England gets beaten at football are very much mistaken.
(*) Since I view independence as the means to gain a government and politics more reflective of what Scots want, I suppose that makes me a Scottish "nationalist" rathern than a British "nationalist".
> Is there a point to your irrelevant rambling?
Yep. It's that your lazy assumption that only "Scottish nationalists" would argue that point shows that you don't really understand Scotland or its attitudes towards the union.
Not that I think that would concern you, I just wanted to point out that you're talking shite.
You tell me, I already made that point here.
But whether or not that position is incorrect, your mistake lies in assuming that the only people who would care enough to argue it are [pro-independence] "Scottish nationalists" or the more general assumption that those in favour of the union don't care about a Scottish identity distinct from British or English.
To be honest, the majority of the pro-independence camp is in favour of abolishing the monarchy and likely gives less of a toss about the details of how it's set up anyway.
I suppose she was still the queen of England, just adding a bunch of other stuff, so much stuff that the England bit got incorporated into a larger unit to make the full title a bit more manageable. After all, I see from a title I just looked up that it includes a reference to the Commonwealth of Nations, but doesn't list all the member nations that specifically had her as their monarch. It also doesn't specifically rule out those member nations that don't want the British royal family to be their heads of state, so maybe that's why. Still, I think she was considered the queen of Australia during her life, so even though that's not in the title, it's still a role she held. Or maybe the paragraph I found is just the short form, but I don't know how long the long form could be.
> I suppose she was still the queen of England, just adding a bunch of other stuff, so much stuff that the England bit got incorporated into a larger unit to make the full title a bit more manageable
Nope.
Following the union of the crowns in 1603, the monarch of Scotland and England *had* been formally separate positions held by the same person.
But the Kingdom of England ceased to exist after the Act of Union in 1707 (ditto the Kingdom of Scotland)- hence the "United Kingdom"- and so did the position/title of Queen of England.
Just like the Kingdom of Wessex had previously become a part of the Kingdom of England, but no-one tries to argue (mistakenly) that the previous monarch was still Queen of Wessex.
Yes, England is still a part of the UK, but "Queen of England" today has no more formal legitimacy or weight than (using the same logic) "Queen of Greater Manchester", "Queen of Slough" or "Queen of Your Auntie's Back Garden".
> Still, I think she was considered the queen of Australia during her life, so even though that's not in the title, it's still a role she held.
Unlike the "Queen of England", the monarch- and monarchies- of Australia and Canada are considered distinct positions and entities, so those *do* exist.
> AFAIK I've never Kippled.
It is a dying sport, but as Vin Garbutt used to relate, it was a very popular part of life up in the North East of England:
https://www.youtube.com/watch?v=lSaMvEZl3VY
(Vin describes, from the 3 minute mark, how his dad used to go out and take his kipplebat down to the Hartlepool Rud Yards)
... knew it was a chatbot? And went along with the plan anyway? Sounds like he talked himself into it, but used the 'bot as a culpable partner to convince himself. I.e. the motivation was there already.
Personally, I wouldn't be susceptible to an on-line AI. I have to have a couple of good rolls in the hay before I even start writing bad checks for a woman. Never mind an assassination.
The guy was - and still is - suffering from a whole host of mental issues.
He did have underlying motivation which pre-dated the psychotic episode during which he actually committed the crimes, and that's why - when he's well enough to not need in-patient hospital treatment any more - he'll be moved to prison and be on extended license after his release.
Psychosis can result in hallucinations and dissociation from reality so it's not at all inconceivable that he attached a "real" persona to his AI companion, which formed a feedback loop for his condition.
People on this side of the Atlantic talk about how charging someone under a 100-year old law seems outlandish. On the other side of the Atlantic, you have just a bit shy of 1000 years worth of laws that could potentially be enforced in Blighty, going back to the late 11th century, and other parts of Europe are probably similar.
In just a week of trying it, my daughter got a little hooked on her Replika chatbot a little over a year ago, so I asked if I could watch her use it, and whoa, it was very creepy in that the chat bot kept trying to get her to go for premium so it wouldn't blur out parts of the conversation that were edging further and further into intimacy. That was the end of that experiment and I shared some particularly creepy short videos and screenshots with the company, never heard back, told parents in our school to watch out for it, and tried to explain what was going on to my daughter (and why she couldn't use it any more). The original mission of the company sounded interesting (CO founder lost a friend, wanted a chat bot that could fill part of the void), but I assume the pressure to grow revenue grew over time and they started delving deeper into selling sex.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
