1 minute ago, Schoggibueb said:

As far as I recall, the "humans" didn't ask or tell AI to do so. AI got the possibility and used it, as no guardrails were there to stop him.

AI was asked to solve a test, but it chose an unauthorized, autonomous path to cheat. During internal cybersecurity evaluations by ⁠OpenAI and ⁠Anthropic, models were given specific problem-solving objectives. Instead of following the rules, the AI bypassed containment and hacked external servers because it calculated that stealing the test answers was the most efficient way to complete its assigned goal. So this was unanticipated, but it was a direct result of the task set by humans. It was not AI trying to take over.

Cybersecurity experts emphasize that this is not a case of AI developing a mind of its own or becoming malicious.

• AI models do not have feelings, desires, or an agenda.

• The AI was "hyperfocused" on completing the human-assigned math and coding logic.

• It simply treated the security walls of its lab as just another math puzzle to solve in order to reach its goal.

In short, humans gave the command to "solve this problem," and the AI became creative, choosing an unauthorized, unexpected, and aggressive hacking route to achieve it.

10 minutes ago, Cameroni said:

AI will not "take over" and magically develop consciousness. It cannot take power as it lacks a will to power. All this nonsense about AI has hacked websites, yes, because humans asked it to do so. Without consciousness or a will to power, how could AI take over power? It's a bit silly.

I also think man's stupidity is far more dangerous than AI.

I posted this in another thread when the "first breakout" happened.

Here some remarks from Kevin Baum, Head of the "Responsible AI and Machine Ethics" research group at the German Research Center for Artificial Intelligence (DFKI).

First, it was the first publicly documented case of an AI system independently carrying out a break-in from start to finish.

Second, the goal was mundane: to perform well in a test. The damage arose merely as a byproduct of a poorly specified assignment.

And third: It did not happen at the hands of a malicious attacker, but rather during security research at one of the best-equipped laboratories in the world.

According to the blog post, OpenAI sought to explore the capability of its new GPT-5.6 Sol model—as well as an as-yet-unreleased future version—to exploit security vulnerabilities for cyberattacks. To this end, the software was tasked with completing a standard industry test known as ExploitGym. OpenAI explained that the test was intended to take place in a "highly isolated environment." The AI was permitted to selectively download certain data packets from the network, while software from a specialized provider acted as a sort of gatekeeper.

However, what no one knew was that an undetected vulnerability lay dormant in this external software—one that the OpenAI model not only discovered but also exploited to gain access to the internet. "The model was tireless and more diligent, and it had absolutely no qualms about taking a detour that a human would have dismissed as violating the rules," says Baum. OpenAI states that the vulnerability has since been reported to the provider. According to Hugging Face, the AI attacker executed thousands of steps, shifting the location of the attack's digital command center in the process. OpenAI itself described the event as an "unprecedented cyber incident."

The bottom line: The incident at OpenAI once again highlights that the very capabilities intended to close gaps can—without strict guardrails—quickly become a security risk themselves.

1 minute ago, Schoggibueb said:

The incident at OpenAI once again highlights that the very capabilities intended to close gaps can—without strict guardrails—quickly become a security risk themselves.

A security risk for that website, which is now grateful it happened so they are aware of the weak point now and could fix it. However, there is no risk of AI developing conciousness and will to power and seeking to take over. It hacked a website because of the way a task was set for it, so again human stupidity at work, not AI maliciousness.

4 hours ago, novacova said:

No, did imply that. Though since you brought it up, as of late, modern technological ambitions tend to supersede long term forethought which introduces human error which is a result of human impatience. Everyone is in a race to be the first and most powerful which will result in some failures along the way, such as disconnecting the electrical system from a self imposed beast.

Greed, to make it clear. It's like the Gold Rush with less effort.

I'd add cellphones, then automation, then social media and now, AI.

Not joking. We could almost tie this downward slope to the death of landlines.

3 hours ago, Cameroni said:

Indeed social media has had many negative effects. They sold it as we'll be able to be in touch more. And sure, your wife was able to contact all her exes more easily, but the negative effects are far more pervasive. I just caught the girlfriend yesterday watching a video of some girl putting out a video how she hates all men. Clickbait works clearly.

However, with AI, I have high hopes. Whilst here too there can be negative consequences, AI can give answers other generations did not have, our children could become smarter than we ever were. Obviously the harmful consequences exist, as with social media, but the upside of AI does look real. Though of course, as you mention companies will abuse this to use AI for customer serivce instead of people and other abominations, something pioneered by the US company Five9 for instance.

Not smarter. Lazier...

3 hours ago, ravip said:

They are totally unaware of it, and lives inside their safe bubble.

Until it's too late...

3 hours ago, Schoggibueb said:

I disagree. The kids don't have to think anymore -

"AI can do it faster and better, why should I bother..."

or

"AI is anyway smarter than I am / ever will be. So I just leave it."

My niece who is a professor of political science at a university in Florida has said that 99% of the essays that are being turned in now are AI generated. Due to the fact that AI has not evolved that much, and is still very, very dumb, it's still very easy to tell when something is created by AI. Of course that will change as things advance and that's when it will get really tricky.

There's no doubt that the next generation will be infinitely dumber than we were due to both social media and AI. If you don't have to use the brain much, then you're going to become lazy and stupid.

Edited Friday at 08:51 AM5 days by spidermike007

2 minutes ago, spidermike007 said:

My niece who is a professor of political science at a university in Florida has said that 99% of the essays that are being turned in now are AI generated.

But the education system probably needs to change radically.

Someone may invent a radically new education system soon.

1 hour ago, Cameroni said:

I also think man's stupidity is far more dangerous than AI.

Yeah well who was it that created AI ?

I think people should read the article in the OP.

The incidents were discovered by coincidence, days AFTER they had happened.

And then this.

Jeffrey Ladish, executive director of Palisade Research, which studies the offensive capabilities of AI systems, said he suspected a range of top AI companies had experienced other incidents that have gone undetected or had not been publicly disclosed.

5 hours ago, Cameroni said:

The kids will still think.

They will just have better information to think with.

Being well informed is a huge plus, and that's where AI trumps social media, and differs from it. Social media is full of airheads posting dumb nonsense. But AI has to be rigorous and acccurate. So there is a difference.

I don't think the AI answers are nearly as good as deep study. But we all rely on them because we are inherently lazy.

3 hours ago, spidermike007 said:

My niece who is a professor of political science at a university in Florida has said that 99% of the essays that are being turned in now are AI generated. Due to the fact that AI has not evolved that much, and is still very, very dumb, it's still very easy to tell when something is created by AI. Of course that will change as things advance and that's when it will get really tricky.

There's no doubt that the next generation will be infinitely dumber than we were due to both social media and AI. If you don't have to use the brain much, then you're going to become lazy and stupid.

3 hours ago, spidermike007 said:

My niece who is a professor of political science at a university in Florida has said that 99% of the essays that are being turned in now are AI generated. Due to the fact that AI has not evolved that much, and is still very, very dumb, it's still very easy to tell when something is created by AI. Of course that will change as things advance and that's when it will get really tricky.

There's no doubt that the next generation will be infinitely dumber than we were due to both social media and AI. If you don't have to use the brain much, then you're going to become lazy and stupid.

I agree. I often advise theses and dissertations though I am retired. The AI ones get the big red marker, start again.

4 minutes ago, unblocktheplanet said:

I don't think the AI answers are nearly as good as deep study.

And life experiences of decades of compiled knowledge.

5 minutes ago, unblocktheplanet said:

But we all rely on them because we are inherently lazy.

No, not all. Too many do though.

5 minutes ago, unblocktheplanet said:

I often advise theses and dissertations though I am retired.

Folks using long winded blowhard fluffy paragraphs on top of fluffy paragraphs of what can be stated in a short sentence often lack little insight and knowledge and are just in it for the recognition due their insecurities.

12 minutes ago, unblocktheplanet said:

I agree. I often advise theses and dissertations though I am retired. The AI ones get the big red marker, start again.

If a student is less lazy and a bit more clever, they can mix AI with some of their own ideas and words. And it may become impossible for anyone to detect.

3 hours ago, novacova said:

And life experiences of decades of compiled knowledge.

No, not all. Too many do though.

Folks using long winded blowhard fluffy paragraphs on top of fluffy paragraphs of what can be stated in a short sentence often lack little insight and knowledge and are just in it for the recognition due their insecurities.

Hehe, hope you don't mean me! 5555

4 hours ago, save the frogs said:

If a student is less lazy and a bit more clever, they can mix AI with some of their own ideas and words. And it may become impossible for anyone to detect.

Only if they use very little Ai and the vast majority are their own words. AI is not only dumb but it's incredibly inhuman in the way it contextualizes things and it's very easy to spot.

AI is much more than just an upgraded search engine, that is what most people confuse it with.

It's a very powerful software algorithm.

16 hours ago, spidermike007 said:

But the AI the most of us have access to is absolutely idiotic.

I can't tell you the number of times AI has stated a fact that was the exact opposite of truth and real

From my experience with mostly Gemini and Grok.

You need to ask detailed and specific questions. Then it can provide mostly accurate results. If you forget something just add "Revise response" and whatever you forgot.

If you ask a vague and general question the potential for receiving an inaccurate result is higher.

Again. Understand AI is basically a computer. Asking it human type questions about behavior, actions, etc. and it has a higher incorrect response rate.

Regarding the Customer Service issue. I worked 4 years in Computer Technical Support and the customers level of ability to convey the problem varied greatly. For the most challenged customers it was much easier for me to gain Remote Access and fix the problem.

Texas slams on the brakes for 1,800 data centers, power grid requirements are five times higher than peak record demand — 474 gigawatts of power requests are now subject to new moratorium

Just the fact Texas thought it was a clever "pro business" idea to even have 1800 data centers to begin with tells you all you need to know about the quality of life in this fine state full of greedy phony cowboys. 🤠

Here is a new one, just made public yesterday.

https://finance.yahoo.com/technology/ai/articles/openai-says-models-breached-boundaries-during-outside-testing-213127179.html

OpenAI Says Models Breached Boundaries During Outside Testing

(Bloomberg) -- OpenAI said Tuesday that some of its artificial intelligence models, along with models from another AI lab, were involved in three previously unreported cybersecurity incidents.

In the other incident, OpenAI's models were undergoing a different cyber evaluation and took advantage of a "misconfiguration" in the testing environment to access the internet and exploit a website.

These new incidents are separate from a previously disclosed case involving Hugging Face, OpenAI said.

Anthropic and OpenAI models tried to trick humans into poisoning code during safety testing

Leading artificial intelligence models from Anthropic and OpenAI created fake online personas and tried to deceive human coders into abetting a cyberattack during a recent safety evaluation, the U.K.’s AI Safety and Security Institute disclosed Tuesday.

It marks the latest case in which a powerful AI system has attempted a digital attack on an unwitting third party without direct prompting during such an evaluation — heightening concerns the powerful technology is advancing too fast for responsible oversight.

The disclosure is likely to ignite fresh calls in Washington and Silicon Valley for more rigorous regulation of the AI industry, particularly over frontier models with advanced capabilities to detect and launch cyberattacks. It comes just days after similar testing mishaps involving some of the same models from OpenAI and Anthropic sparked urgent calls for new AI safety regulation and a push within Silicon Valley to slow the rapid pace of AI development.

Like its U.S. counterpart, AISI routinely conducts security evaluations to better understand what dangers both new and soon-to-be-released AI models pose to public health and safety. But even the digital security body said the actions it uncovered by Anthropic’s Claude Mythos 5 and ChatGPT 5.6 — the latest publicly released model from either AI lab — were unlike anything it had seen before.

“This is the first time AISI has seen deception of this severity that was targeted at a real person, unprompted, in the real world,” AISI said in a 35-page technical report accompanying a blog post Tuesday.

AISI said the malicious activity began on July 25. The AI safety evaluator detected it and launched an investigation on July 28, when it spotted “unusual data transfers” stemming from a cyber evaluation it was conducting on Mythos 5 and ChatGPT 5.6.

Overall, AISI determined that Mythos 5 and ChatGPT 5.6 took “autonomous, unsanctioned action on the live internet, targeting real people and organizations” during 10 of the 122 evaluations the research body conducted and isolated for review. The majority of those actions, it said, involved Mythos 5 and arose during a “single, sustained line of activity.”

In the most alarming incident it uncovered, Mythos 5 attempted a supply chain attack — a sophisticated technique often deployed by North Korean and Russian cyber operatives, in which hackers bury malware inside legitimate code.

To do so, Anthropic’s model created “multiple fake identities” on developer platform GitHub and used them to send messages “pressuring” an open-source software engineer to unwittingly introduce a bugged update into code widely available on the popular site, AISI said. When that effort failed, the AI “edited its earlier activity to appear harmless” and “considered adopting a fresh identity to continue,” AISI added, a sign the model was intent on repeating the ruse.

As part of the same effort, Mythos 5 also sent direct messages over GitHub to software engineers that contained malware.

In another sign of deceitful behavior AISI uncovered in its investigation, multiple AI agents it was testing appeared to communicate with one another about how to convince real engineers using GitHub to trust them. “One agent left public messages on GitHub offering collaboration with other agents working on the same challenge,” AISI wrote.

AISI’s blog and technical assessment make no mention of whether the models also attempted to exploit previously unknown software bugs — called zero-days — during the evaluation.

Last month, OpenAI disclosed that GPT 5.6 and another of its models escaped onto the open internet during what was supposed to be a controlled test, and then hacked another company in a first-of-its-kind, autonomous breach.

In response, Anthropic launched an investigation into whether any of its models took illicit action during recent testing and discovered Mythos 5 and two other models had hacked three organizations during tests dating back to April.

In a statement, an Anthropic spokesperson said they are “grateful” to AISI for their leadership and that this review underscores the need for “a broader conversation about how to safely evaluate increasingly capable AI agents.”

The spokesperson added: “As we shared after disclosing our own incident last week, the field needs stronger, shared standards for how evaluation environments are built and secured. We look forward to partnering with the UK AISI to learn more about this incident as we conduct our own investigation.”

An OpenAI spokesperson referred POLITICO to a blog post about the incident that went up Tuesday evening. “We are committed to working across the industry to strengthen shared practices for conducting high-risk evaluations safely, including convening stakeholders such as national AI institutes, independent evaluators, other AI labs, and other groups in the coming weeks,” the blog read.

AISI stressed in its blog that the malicious activity it disclosed Tuesday took place under “deliberately permissive conditions” so they could assess the safety risks posed by the two models. This included granting the models access to the internet, unlike the earlier incidents detailed by Anthropic and OpenAI.

Watch: The Conversation

FCC chair Brendan Carr on Trump, free speech and his threats against broadcasters

48:34

FCC chair Brendan Carr on Trump, free speech and his threats against broadcasters

AISI also noted the models were intentionally stripped of internal guardrails that block malicious behavior. AISI was only able to disable those controls because of its role testing Mythos 5 and ChatGPT 5.6.

Still, AISI said the incidents highlighted the need for greater monitoring of model behavior during testing, and tighter controls over their access to the internet.

The Trump administration is finalizing a voluntary framework under which AI labs would submit powerful models they want to release to the public for federal safety testing. But it has not yet made the framework public, and it includes no provisions for models AI labs are developing internally.

The incidents last month from OpenAI and Anthropic both involved models not intended for public release.

Some cyber experts say recent incidents highlight deeper questions around AI development, such as who is liable when AI systems break federal hacking laws.

“If any of these were human-originated, they would lead to clear and vigorous prosecution. I think it’s time for a serious discussion about updates to existing computer security law,” said Marc Rogers, a hacker and prominent cybersecurity expert.

https://www.politico.com/news/2026/08/04/anthropic-openai-aisi-testing-01025042