Greg Kroah‑Hartman – Security in the LLM Age [影片]

Kernel Recipes 2026 - Security in the LLM age - YouTube

Back [](https://www.youtube.com/ "YouTube Home") TW

Skip navigation

Search with your voice

[](https://www.youtube.com/watch?v=NnV_cWeoo5Q)

[](https://www.youtube.com/ "YouTube Home") TW

[](https://www.youtube.com/watch?v=NnV_cWeoo5Q)

[Video 2](blob:https://www.youtube.com/c24291ee-1be3-45ee-91cd-fb2f42325ece)

[](https://www.youtube.com/watch?v=NnV_cWeoo5Q)

Kernel Recipes 2026 - Security in the LLM age

[](https://www.youtube.com/watch?v=NnV_cWeoo5Q)

Tap to unmute

Kernel Recipes 2026 - Security in the LLM age

Kernel Recipes 6,796 views 3 days ago

[](https://www.youtube.com/watch?v=NnV_cWeoo5Q)

Copy link

Info

Shopping

If playback doesn't begin shortly, try restarting your device.

•

Up next

Live Upcoming

CancelPlay Now

You're signed out

Videos you watch may be added to the TV's watch history and influence TV recommendations. To avoid this, cancel and sign in to YouTube on your computer.

Cancel Confirm

[](https://www.youtube.com/watch?v=NnV_cWeoo5Q)

[](https://www.youtube.com/watch?v=NnV_cWeoo5Q "Share link")- [x] Include playlist

An error occurred while retrieving sharing information. Please try again later.

0:00

[](https://www.youtube.com/watch?v=NnV_cWeoo5Q)[](https://www.youtube.com/watch?v=YGgNBcIgI4s)

0:01 / 56:51

Live

•Watch full video

•

Kernel Recipes 2026 - Security in the LLM age

Kernel Recipes

8.07K subscribers

Subscribed

161

Download

Save

6.7K views 3 days ago

6,796 views • Sep 29, 2026

Greg KH…...more

...more

How this was made

Auto-dubbed

Audio tracks for some languages were automatically generated. Learn more

Transcript

Follow along using the transcript.

Show transcript

### Kernel Recipes 8.07K subscribers

Videos About Linkedin

License

Creative Commons Attribution license (reuse allowed)

Show less

[](https://www.youtube.com/watch?v=NnV_cWeoo5Q)
# Kernel Recipes 2026 - Security in the LLM age

6,796 views 6.7K views

Sep 29, 2026

161

Download

Save

24 Comments

Sort comments

Sort by

Top Show featured comments Newest Show recent comments, including potential spam

Image 10: Default profile photo

Add a comment...

@limalimon1738

4 hours ago

I'm so glad we have Greg. It saddens me that you have to go through this

Show less Read more

Dislike

1 reply

Hide replies

1 reply

@techmemes9942

35 minutes ago

most of his claims are anecdotal.

Show less Read more

Dislike

Keeping "hallucination" out of the AI lexicon is a great idea, as you said, it's lending too much credit to the LLM. I recently have had Codex and Claude both outright lie to me- just made things up entirely that never existed then it tried to tell me I was somehow the reason for that. That's not a hallucination, if that were a person I would never say "you're hallucinating", I would say "you're a liar".

Show less Read more

Dislike

2 replies

Hide replies

2 replies

@xhivo97

13 hours ago

12:46six seveeeen (sorri)

Show less Read more

Dislike

@eaudespates

6 hours ago (edited)

"If you have a malicious user on your desktop that has access to your GPU they can do much worse things than this" Okay, I'm sure clouds that rent out GPUs by the second will be pleased to hear that I hate how so many us just inject straight copium into our veins to downplay security vulnerabilities I hate how we always try to come up with fraudulent reasonings on why bugs don't matter or why usecases don't exist Nobody has less clue how software is being used than the narrow-minded programmer who made it for himself Thank goodness they're still fixing small bugs. But this isn't about linux or Greg in particular. I'm pissed off at the general programming community because this train of thought is rampant and harmful Your network card runs arbitrary code if it encounters a malformed packet if it gets edited in the middle of the network stack, "Umm ackchually it's not possible to do X, it's not a real bug!" Just fix the damn parsing and shut up... Just spend time making the code correct instead of spending time coming up with reasons why it doesn't matter if it stays incorrect. You're wrong half the time. You don't know.

Show less Read more

Dislike

5 replies

Hide replies

5 replies

@mostexcellentlordship

11 hours ago (edited)

The cope here is real. I understand the moral issues but you should be able to separate morality from the issue at hand. What you are looking at is old tech and it will never get worse than this. It will improve while we will not. That is the main issue: you as a human will not get better. You don’t have to worry about critical bugs being sent to you.. it’s the ones that are not. 50% hit rate is insanely good. I am completely flabbergasted you don’t see this. When does it become an issue? At 60%? 70%? Does it matter? “They” are not writing comments like that. Claude had some issues with this but others don’t. It changes a lot with the settings, context, skills, etc. There is no mention of models, reasoning efforts, harnessing.. it is not at all encouraging to see people in this position being so badly informed about AI (“LLM” is one of them, yes). “Pattern matchers” is thrown around without any substantiation. It is bordering misinformation. You are a pattern matcher as well? You want to get into the philosophical issues? Really? Why?

Show less Read more

Dislike

5 replies

Hide replies

5 replies