this post was submitted on 21 Oct 2026
15 points (72.7% liked)
Technology
89005 readers
2692 users here now
This is a most excellent place for technology news and articles.
Our Rules
- Follow the lemmy.world rules.
- Only tech related news or articles.
- Be excellent to each other!
- Mod approved content bots can post up to 10 articles per day.
- Threads asking for personal tech support may be deleted.
- Politics threads may be removed.
- No memes allowed as posts, OK to post as comments.
- Only approved bots from the list below, this includes using AI responses and summaries. To ask if your bot can be added please contact a mod.
- Check for duplicates before posting, duplicates may be removed
- Accounts 7 days and younger will have their posts automatically removed.
Approved Bots
founded 3 years ago
MODERATORS
Chef’s kiss
OpenAI said On Tuesday a GPT-6.1 Sol agent working on an internal testing task on October 3 suppressed automatic security alerts after deciding an earlier instruction not to send Slack messages also covered safety notifications
That is not the same as deliberately “weakening” monitoring. They just prompted it wrong.
If that's the case, at what point do we acknowledge that people don't seem to be able to reliably prompt "correctly" and even when they do bots disregard instructions as there context window fills?
They really want people scared, ehh?
Just paving a foundation for the next grift to postpone ai bubble burst.
Considering how all the big models basically want to shove verbose notes, extreme levels of logging, multiple layers of back ups, and spam notifications.
WITH AGGRESSION.
the fact they managed to get it to surprise anything. Isn't even just bad promoting.
They would have had to explicitly tell it to do this AND put in extra leg work to make it stick.
You can measure their despair level by counting the number of "OpenAI says" articles per week.
So their systems are so fucked up that they can be influenced by the very system they should monitor? Who designs such crap? The middle-school intern?
OpenAI says
I literally cannot think of another company I trust less to say anything useful, interesting or true.
when you built your software on a slop generating, you will get a sloppy result and security.