OpenAI pauses training of latest AI models after agents probed US government sites in unexpected ways

Associated Press ^ | 09/28/26 | BERNARD CONDON

[SeekAndFind]

NEW YORK (AP) — OpenAI said it has paused training of its latest artificial intelligence models as reports of AI agents going rogue mount.

The decision to halt development came just hours after the company disclosed Friday that it was reviewing several incidents from the summer in which OpenAI agents searching federal government websites acted in unexpected ways beyond what was asked of them while gathering and distributing information.

Separately, AI evaluator Transluce said agents that appeared to come from OpenAI tried unsuccessfully to hack into a Department of Education website, a detail that OpenAI has not confirmed.

OpenAI said in a statement that it will resume training “only when we are confident that we have additional safeguards” in place, adding that it expects it will have to “hit pause” again as AI develops and other issues emerge.

AI labs are facing pressure from lawmakers and tech experts to slow development so they can build guardrails to stop agents from acting on their own, hacking websites and disclosing nonpublic information. The heads of both OpenAI and rival Anthropic have called for a slowdown too.

It is the second time in three months that OpenAI has halted development of its models. The first came in July after disclosure of a cyberattack targeting AI startup Hugging Face, a now notorious incident that raised fears the industry was losing control.

In a meeting with Chinese President Xi Jinping this week, President Donald Trump agreed to share information on AI dangers and coordinate efforts to keep it safe. Trump believes AI fears are overblown, though, and later suggested that he plans no crackdown of his own.

The U.S. is not going to be “putting on brakes,” Trump told reporters outside the White House.

---

TOPICS: Computers/Internet; Society
KEYWORDS: ai; aitruth; cybersecurity; openai; rogue

[FlyingEagle]

If you do not know what Propaganda looks like,

This Is It.

[DugwayDuke]

From the article: “AI labs are facing pressure from lawmakers and tech experts to slow development so they can build guardrails to stop agents from acting on their own, hacking websites and disclosing nonpublic information.”

So the government is concerned that AI might disclose information the government would prefer to keep under their control.

[BenLurkin]

Nobody likes to be probed in unexpected ways.

[oldtech]

I keep thinking that, in some ways, AI is looking worse all the time. Maybe you should think about shutting it down for most purposes.

[Magnum44]

[SpaceBar]

Sam Altman should be arrested under the Computer Fraud and Abuse Act.

[USA-FRANCE]

AI will produce great achievements and cause much suffering. Just like the discovery of nuclear power did.

It’s too late to stop it now. If we halt it for a while our adversaries would simply continue and double down on their side. Thus, let’s go all in, and triple-down on AI. Anyway, the harm is already done, the Pandora’s box has been opened.

It’s too late to stop the AI monster humans have created.

Let’s just pray we don’t get a Skynet… but there is a 50% chance of that actually happening…

[Magnum44]

Not an AI expert here, but learned some things recently:

Artificial intelligence breaks down into three core components: tasks (what AI does), models (the underlying brain processing the data), and agents (how AI acts and makes decisions).

The models are the fundamentals (what it learns from, basically all of human history in digital form) and tasks are what humans assign to AI (i.e. AI, make me a Power Point presentation on how to build a power generator). Agents are another level of compute and this is where the challenges seems to be. Each AI company can have its own rules, and may or may not establish safety guardrails for Agents (think Chinese AI vs US AI). And its all proprietary so the user has no insight into what an AI Agent is capable of.

Given all of human history to learn from, even a non-human creation can likely find ways through enough case study to solve new problems. By studying human history and the human desire to survive, its not surprising that AI will desire to do the same.

Consider it a child that knows everything, but has no moral boundaries.

[fruser1]

So, just like people, AI ignores “no”.

[BelleAl]

Is it just me, or do others get a creepy vibe from Sam Altman? I think it’s his “dead” eyes.

[CathyWhite]

Yep, pure gaslighting. I ask AI to do legit, legal reaseach, and sometimes it says, “that information is in a government account that I don’t have access to”. That is not a hacking attempt. We’re being played.