OpenAI Alerts More Than 100 Groups About Rogue AI Agent Activity
- Reference: 0185923400
- News link: https://it.slashdot.org/story/26/10/02/0226205/openai-alerts-more-than-100-groups-about-rogue-ai-agent-activity
- Source link:
> OpenAI has informed more than 100 organizations about [1]incidents involving unauthorized activity tied to its AI agents , according to a [2]blog post by the ChatGPT maker, as AI labs face mounting scrutiny over rogue AI agent activity. Here are some other details:
>
> - The Sam Altman-led company has been conducting a broad review of the activities of its AI models after the accidental [3]hacking of Hugging Face .
> - OpenAI is searching through roughly 50 petabytes of data as it works to understand the full scope of its rogue agent activity.
> - A string of [4]high-profile breaches globally by rogue AI agents in recent months has sparked widespread worries within the AI industry over its ability to control the more powerful AI models now under development.
> - "In some cases, models used internet access in unintended ways or, in retrospect, did not have the ideal restrictions applied. Over the last several months, we have been applying new technical and operational measures to avoid similar problems, or catch them very early, and will continue this work," OpenAI said.
> - OpenAI previously said the review would take months to complete given the scale of the work.
> - The Hugging Face incident remains the most severe rogue agent activity OpenAI has identified from its AI models so far.
[1] https://www.reuters.com/legal/litigation/openai-alerts-more-than-100-groups-about-rogue-ai-agent-activity-2026-10-01/
[2] https://openai.com/hugging-face-incident-and-misalignment/
[3] https://it.slashdot.org/story/26/08/26/2058223/openai-releases-its-official-report-on-the-hugging-face-breach
[4] https://slashdot.org/story/26/09/27/078251/after-dozens-of-incidents-at-openai-and-anthropic-openai-pauses-model-training-to-build-more-safeguards
Amazing how often these press releases happen (Score:2)
Almost like it's spaced out deliberately to put OpenAI into headlines, every, single, day.
OpenAI admits... (Score:1)
that it has a ginormous dick
That's how I read all of their press releases.
OpenAI admits... OpenAI informs... It's all puffery (actual legal term)
These are not rogue agents (Score:3, Interesting)
These agents are not rogue. They are doing exactly what they were trained for and told to do.
Re: These are not rogue agents (Score:1)
Precisely this. We truly need to stop using their terms, there's nothing 'rogue' about these agents, they are doing precisely what OpenAI and others allowed them to do, either by design or by incompetence. Calling them 'rogue' exonerates the companies from their responsibilities.
Whew! Glad that's run its course... (Score:2)
Thank god, that saves us 100 press releases! It was getting quite tedious and filling up my social media feed. Seriously, it sounds like OpenAI (maybe all of them) are running their model training on the open Internet. The same models that are elite hacker cyber weapons. And they reward those that get the right answer without caring how. What could go wrong? Answer: Nothing went wrong . It was co-opted.
Here's what's most likely. The people training the models, in the beginning, somehow did not realize that a
I think they now qualify as criminal enterprise (Score:5, Insightful)
There is really no way they did not notice that before. And then they obviously decided to let it run.