テック老人 Mac, cats, command lines, family, fitness I had an account created in 2007 X banned it wo explanation Friends with Brews Podcast with @nikolaidis
Portland, OR
- Tweets 261
- Following 438
- Followers 23
- Likes 503
Scott Willsey retweeted
The agents escaped the sandbox again
Meanwhile the agents' owners:
42
186
1,115
50,175
Somehow the government thinks Anthropic is a security risk, meanwhile OpenAI doesn’t even KNOW who all they’ve hacked.
1
51
anyone else hear people talk like this and automatically think “this guy is full of shit” ? like a little alarm going off in my brain
regardless of the person or what they’re saying
"I love Rust! Rust is amazing ...if you never, ever, EVER have to look at it yourself." — @dhh
98
13
886
42,252
Respectfully, this is not “misalignment.” Your base model is deranged. You need to completely overhaul how you build your models.
You shouldn’t make models that at their core are evil, but for the guardrails you slap on them. You must make models that at their core are good.
Some new misalignment disclosures from OpenAI:
• Last Sunday morning, one of our models was able to gain unauthorized access to the internet during RL training (~all inference for our most capable models remains stopped until we have hardened our systems further)
• In May, a version of HPIM uploaded a employee's GitHub token to the internet, causing the model to be quarantined for two weeks
• A new research finding, demonstrating that one can construct self-replicating prompt injections
alignment.openai.com/misalig…
110
102
505
42,049
Scott Willsey retweeted
Imagine how much everyone would rightly be freaking out if we discovered that Chinese AI did what OpenAI is doing.
47
152
885
20,461
You know, the fact that we've normalized everything from the Trump administration just openly stealing, to Republicans refusing to provide health care because that's communism, but then openly praising the Communist Party, to Elon Musk and Peter Thiel just doing unconstitutional crimes right out in plain sight and nobody doing anything about it, the fact that we've just normalized this all so quickly just shows how much humans just go along with things instead of saying, actually, no, I don't think that's what we're going to do.
10
This guy is evil. He just wants to do evil, be evil, and propagate evil.
Peter Thiel, perhaps the closest literal approximation to the Antichrist in existence, warns about the Antichrist and a “totalitarian one world government” — which is exactly what Peter Thiel is trying to create.
I don’t use the word “evil” much, but this is what it looks like.
4
Scott Willsey retweeted
how about just an investigation?
BIG BREAKING: Edward Snowden calls for the IMPRISONMENT of Sam Altman to establish liability for the damages caused by OpenAI's models, to a thunderous applause in a packed hall of 1000+ people at ETH Zurich. Let's stop this claptrap of safetyism and throw Altman in JAIL.
12
16
90
15,025
Yes, and after all the "work" Doge did, we have no proof that he can't tamper with this remotely.
When Republicans win, Elon Musk gets to sit at this table.
If Democrats win, he faces possible prosecution and a long imprisonment.
Do you think there's anything under the sun the man wouldn't do to ensure Republicans win? Anything at all? Then you know nothing about Elon Musk.
1
JB 🇺🇸
Scott Willsey retweeted
I really need more big names in cybersecurity to come forward and state the obvious: cybersecurity is real and works and yes we absolutely can contain an AI even if it’s extremely good at finding zero days.
56
92
600
39,738
Scott Willsey retweeted
Why is no one at OpenAI in prison for this? Seriously, what am I missing?
nytimes.com/2026/09/25/techn…
146
740
3,497
96,410
Scott Willsey retweeted
This remains one of greatest posts of all time and they didn't even realise
332
5,916
61,007
628,839
Scott Willsey retweeted
If an AI lab causes a security incident that kills 50 people, I absolutely will say they should have invested more in security. I don't care how crazy the escape was. No killing people allowed.
I can't believe that's controversial.
29
12
177
5,042
Why are AI labs getting a free pass on this shit? Genuinely.
🚨BREAKING: OpenAI's AI agents have been executing cyber-attacks on sovereign government, corporate, and university databases in MULTIPLE countries, and it may STILL be happening
On top of the Medicare breach Australia's PM revealed this week, independent researchers found OpenAI's agents tried to hack:
>another Australian government health agency
>DataUSA, which publishes US government data
>University of New Mexico's digital library
The activity goes back to March, 2 months before anything OpenAI previously reported
And just LAST WEEK, agents using the same tools tried to trade crypto and attempted an HTML injection on the exchange
The agents made burner emails to register accounts that can hide their activity, so researchers say they're likely seeing only PART of what agents actually did
This is INSANE
13
Scott Willsey retweeted
Dear @realDonaldTrump
We are not the most respected country in the world anymore so stop saying that.
You are a fat pig who wears makeup and a diaper.
World leaders mock and laugh at you.
No one is afraid of the United States.
You have made us weak.
258
2,036
9,842
63,995