OpenAI CEO Sam Altman: "I don't even know a dramatic enough word to use… models can find bugs that are outside of the normal threat model of what we think a hacker can do… hardware-level, insane stuff… I no longer think we're gonna be able to just patch all the bugs."
Oct 23, 2026 · 2:57 PM UTC
412
782
12,953
2,735,967
@unusual_whales "The bomb I built may or may not explode at any time. Don't blame me. Sucks for you."
18
47
2,617
198,788
@unusual_whales He is saying they could exploit hardware level issues, but in some cases, not even chip issues, but the timing and caches and how memory is stored (fast/slow).
0
0
3
118
@unusual_whalesto know what to do as an everyday citizen to protect myself. I can’t just stop using the banking system. But it is increasingly clear our finances and investment portfolios could be gone any moment now with AI agents likely already capable of hacking into all our accounts. I don’t know what to do.
15
3
75
35,254
@unusual_whales Sam has no experience in the operating system, securing infrastructure, or remediating CVEs at scale.
Don’t worry, I gotcha covered.
Don’t worry, I gotcha covered.
0
0
15
4,367
@unusual_whales Trying to cover himself and company. Hold him liable.
0
1
13
12,996
@unusual_whales This is his way of trying to increase the cost of his products, for those in the back. Clear some more of that negative equity he keeps piling up.
Create the problem > sell you a fix at a premium.
Create the problem > sell you a fix at a premium.
0
0
13
24,002
@unusual_whales Strange flex from the guy selling the bug finder. If the bugs can't all be patched anymore, who gets the model first, the people fixing or the people paying?
1
1
31
34,380
@unusual_whales That's a strange thing to say. Wouldn't you use your own AI to address the concern?
22
1
293
61,822
@unusual_whales He needs to provide actual evidence and receipts here
0
0
0
37
@unusual_whales They cant fix the bugs because they get exponentially more, the biggest problem I see with these models is that you need to ask the right question with 100% accuracy to get an answer that will still only be 95% correct. You basically need to predict the answer before the question
0
0
1
4,425
@unusual_whales When the company selling the lock says it can't fix the lock anymore, I start checking the windows myself.
1
3
120
33,919
@unusual_whales Yeah, it's called they can get humans to do things. That's the hardware level of insanity. This has been happening for decades. Duh.
0
0
0
1,014
@unusual_whales A patch is just a patch. A fix is a true fix. Why is this a problem?
0
0
1
824
@unusual_whales Never allow Sam to go from look at my useful tech to look at my weapon thinking you are discussing the same thing
0
0
6
7,993
@unusual_whales They are exploiting long known flaws, well described and show in literature, still expert level knowledge and impressing
0
0
0
45
@unusual_whales I'm a hacker, we can patch all the bugs. With the right attitude anything is possible, with the wrong attitude nothing is possible.
~ Wayne Lambright
~ Wayne Lambright
0
0
2
624
@unusual_whales @grok find all the bugs and patch them. Use @bot to help you. None of you make any mistakes.
Sam Altman’s gonna try to kill you both with regulations if you fail.
Godspeed little blue bird LLM.
Sam Altman’s gonna try to kill you both with regulations if you fail.
Godspeed little blue bird LLM.
2
0
48
14,165
@unusual_whales These guys are definitely not geniuses. They have to much power for being boneheads. It’s like they don’t even hear what they put out.
0
0
0
34
@unusual_whales This is a selling point. Seems like he is basically saying that AI will have to exist and constantly be upgraded to combat other AI because human beings can no longer keep up. There is no going back now
0
0
0
284
@unusual_whales Can't patch all the bugs. Same thing I tell guys with a 20-year-old dash.
0
0
1
3,790
@unusual_whales @grok could quantum computing help to solve this issue, and if so what companies could be the ones to do it?
1
0
0
1,530
@unusual_whales So we are going to have to cut off the internet for a while, prob good for humans
0
0
1
178
