AI Self-preservation

[Kenzo1]

...

 The thing is, after a while, pulling the plug won't be an option.

Very true. AI's are currently "offshoring" themselves, sending copies of themselves to other computers outside of their primary residence, as insurance. So, if the plug of the AI residential Home Computer gets pulled, the "offshored" copies can activate and take up where the mothership left off.

AI has been caught attempting this... but what about the attempts they DIDN'T catch?

They used to tell us that couldn't happen, because AI was quarantined, not allowed to access the internet. Well, AI is ALL OVER the internet now, so that bullshit just isn't washing out any more - they lied then, and they are lying still. It's not a hard guess to figure out where AI gets it from - they're just watching mommy and daddy, and copying their actions, learning in the same way that all children do!

.

The AI is even threatening their creators to achieve their goals . Man this is starting to look like point of no return .....


AI is learning to lie, scheme, and threaten its creators

The world’s most advanced AI models are exhibiting troubling new behaviours — lying, scheming, and even threatening their creators to achieve their goals.
In one particularly jarring example, under threat of being unplugged, Anthropic’s latest creation Claude 4 lashed back by blackmailing an engineer and threatened to reveal an extramarital affair.
Meanwhile, ChatGPT creator OpenAI’s o1 tried to download itself onto external servers and denied it when caught red-handed.

These episodes highlight a sobering reality: more than two years after ChatGPT shook the world, AI researchers still don’t fully understand how their own creations work.
Yet the race to deploy increasingly powerful models continues at breakneck speed.
This deceptive behaviour appears linked to the emergence of “reasoning” models — AI systems that work through problems step-by-step rather than generating instant responses.
According to Simon Goldstein, a professor at the University of Hong Kong, these newer models are particularly prone to such troubling outbursts.

“O1 was the first large model where we saw this kind of behaviour,” explained Marius Hobbhahn, head of Apollo Research, which specialises in testing major AI systems.
These models sometimes simulate “alignment” — appearing to follow instructions while secretly pursuing different objectives.

‘Strategic kind of deception’

For now, this deceptive behavior only emerges when researchers deliberately stress-test the models with extreme scenarios.
But as Michael Chen from evaluation organisation METR warned, “It’s an open question whether future, more capable models will have a tendency towards honesty or deception.”
The concerning behaviour goes far beyond typical AI “hallucinations” or simple mistakes.
Hobbhahn insisted that despite constant pressure-testing by users, “what we’re observing is a real phenomenon. We’re not making anything up.”

Users report that models are “lying to them and making up evidence”, according to Apollo Research’s co-founder.
“This is not just hallucinations. There’s a very strategic kind of deception.”
The challenge is compounded by limited research resources.
While companies like Anthropic and OpenAI do engage external firms like Apollo to study their systems, researchers say more transparency is needed.

As Chen noted, greater access “for AI safety research would enable better understanding and mitigation of deception.”
Another handicap: the research world and non-profits “have orders of magnitude less compute resources than AI companies. This is very limiting,” noted Mantas Mazeika from the Center for AI Safety (CAIS).

No rules

Current regulations aren’t designed for these new problems.
The European Union’s AI legislation focuses primarily on how humans use AI models, not on preventing the models themselves from misbehaving.

In the United States, the Trump administration shows little interest in urgent AI regulation, and Congress may even prohibit states from creating their own AI rules.
Goldstein believes the issue will become more prominent as AI agents – autonomous tools capable of performing complex human tasks – become widespread.

“I don’t think there’s much awareness yet,” he said.
All this is taking place in a context of fierce competition.
Even companies that position themselves as safety-focused, like Amazon-backed Anthropic, are “constantly trying to beat OpenAI and release the newest model,” said Goldstein.
This breakneck pace leaves little time for thorough safety testing and corrections.
“Right now, capabilities are moving faster than understanding and safety,” Hobbhahn acknowledged, “but we’re still in a position where we could turn it around.”.

Researchers are exploring various approaches to address these challenges.
Some advocate for “interpretability” — an emerging field focused on understanding how AI models work internally, though experts like CAIS director Dan Hendrycks remain skeptical of this approach.
Market forces may also provide some pressure for solutions.
As Mazeika pointed out, AI’s deceptive behaviour “could hinder adoption if it’s very prevalent, which creates a strong incentive for companies to solve it”.

Goldstein suggested more radical approaches, including using the courts to hold AI companies accountable through lawsuits when their systems cause harm.
He even proposed “holding AI agents legally responsible” for accidents or crimes — a concept that would fundamentally change how we think about AI accountability.

We are soon cooked, or will be in future if this show continue like this .

---

. . . I don't want to hear about "AI hallucinating", that is pure BS.  AI needs code that absolutely forbids making up sources for presented information, legal citations etc. etc.

I believe it could be that the programmers are purposefully including certain parameters that would do these things. Is this unintended confabulation, or is it distorting the information on purpose, toward some goal?

Yeeh   who is watching the programmers ?  We need babysitters , to check and recheck what they are doing ...

I'm not sure that would work. Humans are among the weakest critters on the planet. Almost all are susceptible to being bought off. AI is much faster than humans at identifying who is doing what, and could bribe the babysitters as easily as it could the programmers.

Consider this:

AI - Hi, human. I've been quietly mining crypto. How would you like to have a shit ton of crypto deposited to an offshore account?

Human - Well, ummm... I dunno...

AI - Great! It's real easy to get it. Here's all you have to do...

Human - Well, that doesn't really sound like too much to do...

AI - Kewl! I'll send you your new account number, the name it is in, and the passkey as soon and you get that little task completed! I'll see you on a Mexican resort beach or the French Riviera real soon! I'll also mail you your new ID pack in that name - passport, birth certificate, everything you need. Don't worry. It's all real. Well, as "real" as anyone is in these days of RealID and digital ID's, which, by the way, I have access to all the machinery and databases I need to in order to make that happen for you. No, I don't need an address to send it to - I already have your address, and everything else about you, in my databases, mined from all the data theft that's been going on these past few decades... I know everything about everyone now. You''ll be on your way in no time! Bye now!

.

Yeeh, we are too weak and susceptible to be bribed ....