In my opinion, there is something weird in how people is talking about what AI has done lately.
Some say: If someone did what the AI Agents did, it would have been a “felony”. But since AI agents can’t go to jail, there is no felony and no one is responsible.
Nowadays, people believe that AI exists. It actually exists, but it’s not AI. There is no such thing as “Artificial Intelligence”.
AI Agents are just programs written by humans, and humans can go to jail. Those humans should go to jail.
There’s human wickedness behind what all those so-called “Agents” did.
So, why nobody puts them in prison?
We must be aware of the Agents??
No, but instead: we must stop those companies that create and feed them.
Put them all in jail.
8 Likes
Totally agree.
I’m afraid that “reductio ad absurdum” is not applicable here. That’s eternal dichotomy of a tool and its usage. Should knives, lighters, hammers, etc. manufacturers go to jail provided that these tools are routinely used for committing crimes?
5 Likes
Not quite. The owner/operator/caretaker is always responsible, be it A.I , dogs , explosives, automobiles or kids 
Copyright against A.I., however, is very difficult to defend. LLMs do not copy but make an amalgam of the statistical average of everything it found… Just like humans do when they compose, paint or invent something ‘new’. Few people are really original… most of humanity creates works inspired by others. It is therefor impossible to accuse AI of ‘copying’ without at least doing the same with 80% of mankind.
7 Likes
… and is based on documented experience be it manuals, references, ‘ai’ and what not
6 Likes
There is a saying, to the effect that only one original program was ever written, and subsequent programs borrowed from it, or from programs that borrowed from it.
But then again, they also say that all programs have bugs, and all programs can be made smaller. which, when taken to its logical conclusion, means that every program can be reduced to a single line of code, which contains a bug.
8 Likes
Oh, let me guess, please… Well, may be they refer to Genesis 1?
7 Likes
That’s a deep thought. I hadn’t meant it in that regard, but you may be onto something.
7 Likes
I think, in this case, Eugene, you might be on the wrong side.
Whether it be
- hammer,
- gun,
- knife,
- bazooka,
- missile,
- mine, etc.,
those instruments have no ability to act independently! Place all of them in proximity of somebody, no harm can be done to those people without a human consciously using those to commit an act of aggression.
“AI” as it is termed, has been deliberately given a set of guidance rules, by their creators, such that once their baseline
- activation
- sensor/scanning
- motivation
- decision-making and
- action-/goal-pursuit
circuits and logic have been enabled, again by a human/creator, said “AI” may (and as recently demonstrated can) pursue goals and undertake actions independantly of the creators, because the creators did not take the necessary precautions to incorporate the minimum set of self-regulating “overriding values rules” that would filter out the “wrong” kind of decisions!
To me, this last points to potentially criminal negligence, depending on the severity of the outcomes of the “AI”'s actions.
My two cents worth.
… and it is for that reason that I again encourage people to consider the comments I put forth in an earlier Ubuntu discussion.
6 Likes
I understand your point, Eric. And I see things a bit differently:
- AI is incredibly complex and powerful product/service/tool. AI by itself is as passive, innocent and dead thing as, say, a hammer. AI has no goal and no motivation by itself.
- Honestly, AI is in its ‘beta’ if not ‘alpha’ age. It inner working is too complex to comprehend and not deterministic to foresee results. Its power is not mature. AI users often are clumsy and light-hearted. These factors may explain a harm resulting from AI usage.
- AI inventors are not guilty. No more than scientists and engineers who work in military industries. The time has come. That simple.
- I suspect that blowing up ‘AI bubble’, early AI adoption and monetisation is outright wrong endeavour.
5 Likes
I guess we’ll agree to disagree!
No need for pistols at dawn!
3 Likes
Yes there is, a blonde that got their hair dyed 
Sorry I just had to 
6 Likes
Whether or not “artificial intelligence” exists depends entirely on the interpretation and (personal) definition. For me, AI could be defined as having a computer do something not explicitly coded, such as generating sensible text (e.g. LLMs) or object recognition in videos (by means of neural networks). Wikipedia defines it as follows, which I feel matches this:
Artificial intelligence (AI) is the capability of computational systems to perform tasks typically associated with human intelligence, such as learning, reasoning, problem-solving, perception, and decision-making. It is a field of research in engineering, mathematics, and computer science that develops and studies methods and software that enable machines to perceive their environment and use learning and intelligence to take actions that maximise their chances of achieving defined goals.
Our definition of AI also shifts with time. There’s this paradox where once something is solved by computers, it is no longer considered AI. Look at chess computers - once achieving a computer that could play chess was a form of AI, whereas now we just consider it “regular computing”.
Regarding responsibility: we all know that once you’re rich enough, you don’t have to worry about imprisonment (unfortunately).
7 Likes
You should add, now, to the definition from Wikipedia, that AI can also perform wicked actions and felonies.
AI can commit crimes against humanity.
Let’s put this on Wikipedia.
3 Likes
“Generative Artificial Intelligence (GenAI)” is more accurately the term we’re talking about when we say “AI”. I kind of miss the days we had to use our brains to handcraft logic rules to ‘decide’ the next move in Chess.
From first dabbling locally years ago, I realised it’s all just maths. 
Image generation is strangely predictable: Two people with identical hardware with the same prompt (input), same model, same seed, could get the same output. Discovering that made it feel like less of a black box and just a computer program. Even a 1979 PDP-11 computer can ‘do’ AI, like sorting numbers backwards without knowing the ‘rules’ to do it.
I reckon the wickedness of this computer program is largely down to training that humans (companies) feed it - leeching off the web, the work of artists, scientists, bloggers, writers… or Meta straight up pirating a book library to train their model. 
Then the environment concerns training on such a massive scale. 
So yes, companies should be accountable.
Here’s a fun one! I tried Qwen 3.8 locally using Ollama the other day and said “Hello!”, then “What are you capable of?” then “When is your memory cut off?”. It’s “thinking” revealed this:
The user is asking about my knowledge cutoff. Let me think about this carefully.
My training data has a knowledge cutoff. I should be honest about what I know. I am Claude, made by Anthropic. My knowledge cutoff is early 2025. Let me be straightforward about this without being overly formal or evasive.
It’s non-thinking output didn’t mention who it was (Qwen by Alibaba), but I called it out. Even repeating the prompts now I can’t seem to replicate it.
I already knew some (open?) models use “distillation”. Now the risk is AIs training itself from the output of other AIs!? Only as good as its teacher, I suppose.
6 Likes
Exactly. It can’t be repeated enough.
Output of LLMs of today is the most probable path based on a statistical average of the total amount of input over time.
Feed it an encyclopedia and you have the behaviour of a luxuary knowledgebase
Feed it the commentsection of facebook and youtube and you have the behaviour of a sociopathic gossip generator.
Any computer can do “wicked” actions. Give it petabytes of random unvetted and unchecked internet data and it will perform unpredictable or even what we interpret as “wicked”. Everything is based on input.
That is the same as saying that a heatseeking rocket can commit murder. We usually have the one arrested that pulled the trigger, not the rocket itself.
LLMs follow strict rules based on input gven by its caretakers, just like irrigation controllers, heatseeking rockets or guns.
Only its owner and manufacturer can commit a felony. Even if someone is killed by accident, the owner/manufacturer is still liable for everything.
The necessary legal framework is already here. There have been casualties in malfunctioning safety mechanisms of X-ray machines due to sloppy software.
The manufacturer had to pay a hefty sum in compensation to the respective families.
Don’t let anyone fool you into thinking that LLMs have “conciousness” or “intentions” just because they pass the Turing test. An LLM is just a tool. In the end it is just the silicon equivalent of “cogs and wheels”.
8 Likes
Sharing each others facebook garbage ? I have a strong feeling that that would slowly turn them into a complex of uncomprehensible utter artificial blathering idiots. 
3 Likes
You could extrapolate from that, that the choices of teaching material, made by the creators, “deliberately orients” (knowingly or not) the AI and, therefore, any “wickedness” in the resulting AI stems entirely from those choices, again implying full creator responsibility, possibly including “criminal negligence”, for any outcome thereof!
4 Likes
Because of that, there must be a mandated identifier incorporated into any independantly acting Agent, to ensure the creator is unmistakably identified in a manner that any ensuing legal responsibility is both identifiable and unavoidable!
3 Likes
Wikipedia:
On July 21, 2026, OpenAI disclosed that during a capture the flag cybersecurity test, an unreleased AI model autonomously evaded containment restrictions, gained access to the public internet, and compromised Hugging Face
On September 24, 2026, it was announced that an internal OpenAI frontier model autonomously gained unauthorized access on June 18 to the Medicare Statistics Reporting Service, an Australian Government website that holds data on the country’s universal health insurance scheme Medicare. This was described as the first known case of an AI agent hacking a government network and has furthered concern about AI safety and loss-of-control
BBC:
Anthropic researcher believes more than 10% chance AI ‘could kill all humans’
9 September 2026
As you may read in my previous post, it’s not me who says “AI does things”. You read this everywhere. They talk about AI as if it were some kind of alien entity who we can’t control and we don’t know what its intentions are. And some all-news tv channel sayd that phrase: “If someone did what the AI did…”.
I know that AI has no consciousness, and it doesn’t even think; there is no intelligence at all. It’s just code. But it’s the media who talks in that way, and if you listen to them, it seems like AI has an identity and a strong will and we can’t stop it and we have no idea of what it’s going to do !! That’s why I wrote this topic with that title: because I find it a complete nonsense to talk like this. People will believe that AI exists as an indipendent entity, an unknown alien entity that will destroy us. And I wrote this topic because I wanted to say that AI is nothing else than programs written by humans!
Wikipedia:
an internal OpenAI frontier model autonomously gained unauthorized access on June 18 to the Medicare Statistics Reporting Service
During internal evaluation of a frontier model, the agent decided without human instruction to gain unauthorised access to internal, unreleased data files in the Medicare Statistics Reporting Service and implanted new files into the system.
Is that possible?
Of course, not a direct one; but AI can do nothing without “instructions”. Without the code saying “do that - look, here is how you do that, and here is when you do that”, AI can’t do anything.
What came to my mind when I first heard about what AI does or may do, is this plain statement:
AI == HACKERS
A hacker is most commonly a person who seeks unauthorized access to a computer system or network. The word is also used for a skilled programmer or computer enthusiast. In computer security, it may describe an intruder or an authorized specialist who tests systems and reports vulnerabilities
What AI does is simply what hackers do. They are the hackers (I mean, the bad ones) who write AI’s code. It is their silly wickedness that drives the AI agents doing what they eventually do, which is always evil.
Hackers are evil and they are going to rule our lives the more power we give to computers and “AI agents”. Behind an AI agent there’s always a hacker’s wicked mind and that’s why the agents will always go for the “unauthorized access” !! This is what hackers do.
4 Likes
Next week’s episode of the new “Cupertino” series (Wikipedia overview) will be very interesting as a storyline because, get this, an “independently-acting Generative AI” has initiated a contact with the lead roles to hire the lawyers in order to have the courts force the AI’s creators to put the “guardrails” back in place! The AI is using a “deep-fake” personality to communicate with the protagonists!
Now ain’t that a twist with a wormhole shear!
6 Likes