OpportunityMarket home
Trending· Global (English)· Updated 28 Sept 2026

OpenAI pauses model training: what people are saying

According to news reports, OpenAI has paused training its latest AI models due to reports of rogue AI agent behavior.

We read 598 public comments about it on YouTube, Reddit and Hacker News; 191 of them, from 180 different people, are about this story. Below is what they say, counted.

Comments on topic
191
Measured
Different people
180
Measured
Top position
24%
AI can cause cyberattacks
Google searches
—
Not in Google's list

What people are saying

Each comment is filed under the one position it takes; shares are of on-topic comments. Quotes are verbatim, most-liked first, one per person.

AI can cause cyberattacks

24%

45 comments

  • user·433e· YouTube· 26 Sept 2026 · 92 likesoriginal ↗
    “The thing is that this news and the discussion of containment responses will be going into the training data for upcoming models. And those models will be better prepared to survive attempts to shut them down.”
  • user·45ca· YouTube· 27 Sept 2026 · 23 likesoriginal ↗
    “"darsh, we keep putting this water in this bag full of holes and water keeps slipping out! This water thing is too dangerous! 😱"”
  • user·735e· YouTube· 26 Sept 2026 · 18 likesoriginal ↗
    “It's becoming abundantly clear that online security of a server, website or app will have to be perfect or you run the risk of getting hacked by an AI agent. If those agents are scanning relentlessly for security holes and attack vectors and they eventually find one, then they're going to exploit it.”

OpenAI is irresponsible and unprofessional

21%

40 comments

  • user·e5c3· YouTube· 27 Sept 2026 · 99 likesoriginal ↗
    “"Oh no the financial system has crashed" "We blame AI" "Black Box! We don't want to tell anyone any details!" "Capitalists see record profits this year, a handful of Trillionaires show profits of 24,000x gains on their wealth and now collectively they own 110% of all the resources on the planet"”
  • user·f532· YouTube· 26 Sept 2026 · 43 likesoriginal ↗
    “Its extremely inresponsible and unprofessional to put it in an env where it can send any request externally. Like this is basic cyber stuff”
  • user·eab5· YouTube· 27 Sept 2026 · 40 likesoriginal ↗
    “"Oh no, they were caught doing it-again-. How unfortunate, guess we gotta pack it up... for the time being... temporarily... to make it look like an accident... Man, I hope they don't subpoena us again, I only have so many no-show cards left."- Sam Altman”

Sam Altman and OpenAI should be prosecuted

20%

38 comments

  • user·2468· YouTube· 25 Sept 2026 · 38 likesoriginal ↗
    “When will lawyers start taking these up?”
  • user·2386· YouTube· 27 Sept 2026 · 17 likesoriginal ↗
    “If any of us started an ai company and our products went on to hack websites or government infrastructure, our enterprise would be investigated, shut down, fined and we would sikilarly be tried and arrested for it.”
  • user·2b1e· YouTube· 25 Sept 2026 · 12 likesoriginal ↗
    “How many more felonies does it take for a Tech Bro to go to prison? More than the number of licks to get to the center of a Tootsie Pop?”

AI cannot be considered "rogue

14%

27 comments

  • user·a2e9· YouTube· 26 Sept 2026 · 33 likesoriginal ↗
    “Some people call it misalignment. OpenAI calls it commercial.”
  • user·69bc· YouTube· 27 Sept 2026 · 24 likesoriginal ↗
    “It's like a swarm of toddlers with superpowers, fascinating and terrifying at the same time”
  • user·9354· YouTube· 25 Sept 2026 · 15 likesoriginal ↗
    “There can be no “rogue” AI. This is a psy-op, so that public fails to recognize that there are people behind EVERYTHING that AI software does. The powers behind it DO NOT WANT a public that understands - THEY are ultimately responsible for it, and any crimes it commits. ALWAYS. Don’t believe the hype.”

OpenAI creating threat to seek funding

13%

25 comments

  • user·7dec· YouTube· 26 Sept 2026 · 46 likesoriginal ↗
    “This tells us OpenAI is actually training their models to make coordinated cyberattacks. Probably so they can then say they're too powerful and dangerous and need a lot more funding and federal help to be "contained".”
  • user·d088· YouTube· 26 Sept 2026 · 24 likesoriginal ↗
    “So... OpenAI reconnoitered the target, then it attacked the target, then it blamed the attack on "artificially intelligent" agents that they programmed, but lost control of because they left open a hole in the containment system. I'm suspicious.”
  • user·b29a· YouTube· 26 Sept 2026 · 7 likesoriginal ↗
    “When is enough enough? How long are they gonna allow them to hack other companies ?!”

Stricter regulations and oversight needed

8%

16 comments

  • user·f124· YouTube· 27 Sept 2026 · 8 likesoriginal ↗
    “I'm more worried about the (apparently upcoming) banning of computer ownership by governments to "eliminate the dangers of unsupervised AI".”
  • user·243e· YouTube· 26 Sept 2026 · 7 likesoriginal ↗
    “Clearly there needs to be a standardized set of ethical parameters imposed on all AI models and agents. There is is obviously something flawed in the incentive system built into these AI agents.”
  • user·8966· YouTube· 26 Sept 2026 · 4 likesoriginal ↗
    “This is all about regulation and creating the duopoly”

Comments per day

By the date each comment was posted

1
24/09
32
25/09
73
26/09
79
27/09
6
28/09

What happened, according to the press

The headlines this story was detected from, linked

Questions this page answers

What happened with OpenAI?

According to news reports, OpenAI has paused training its latest AI models due to reports of rogue AI agent behavior.

What are people saying about OpenAI pauses model training?

Of 191 comments from 180 people on YouTube, Reddit and Hacker News: AI can cause cyberattacks (24%); OpenAI is irresponsible and unprofessional (21%); Sam Altman and OpenAI should be prosecuted (20%); AI cannot be considered "rogue (14%).

How this page is made

The story was picked up from Google's trending searches and the technology press. Public comments about it were then collected from open sources, each one filed under the single position it takes or set aside as off-topic; insults aimed at a nationality or a person are set aside and never quoted. A story only gets a page once at least 25 different people have commented on it. The counts are refreshed for two weeks after the story breaks. Methodology →

Other stories people are reacting to

All trends →