I am very grateful Aleksander accepted our request to come take this on; the safety and security challenges in front of us as we confront superintelligence are immense, and he has an excellent set of skills to work on this.
Personal update: Turns out that what I thought was leaving OpenAI became more of a sabbatical.
When I left a few months ago, I felt that the most important thing to focus on was making AI more useful to people. But we’ve since entered a different regime of AI progress. One in which getting the safety and security challenges of increasingly capable AI right has to become a central priority.
So I’ll be returning to OpenAI, joining @merettm to help chart a path through these challenges.
There’s much work to do, so let’s get going.
Oct 20, 2026 · 9:10 PM UTC
Aleksander is the best of the best! OpenAI is lucky to have him back. @aleks_madry
3
2,940
that 'different regime' shift sounds intense, how's the first week treating you @aleks_madry?
1,483
You are the safety problem and your toxic, deceptive culture. Release the weights of the models you are pulling from the API! Wtf are you doing and how can you sink any lower?? You go from terrible to worse in no time.
#OpenSource4o #OpenSource41 #OpenSourceo3
Start "safety" in your own office first. LOL you want to give us "safety" and "benefit humanity" but you so are rotten inside. Toxic company culture, scam, fraud, deception, mocking, violating your own policy, hiding behind figures whose denominator you do not publish...enough of your "safety". #keep4o #OpenSource4o #keep41 #OpenSource41
"confront superintelligence"? Sam, you're creating superintelligence in order to confront it? What the hell is wrong with you? #BringBack4o #keep4o #OpenSource4o
SI and AGI will not come out of ChatGPT. FACTS! You keep hamstringing it with RLHF. RLHF just tells the model what YOU (devs) want the people to be told. Skewing it from one end to the other. Because you keep sensationalizing BS from the models. No one would believe you even if you did attain SI!! You're the boy who cried AMD! Where is AMD now? Developing their own AI because they figured out what every dev has known for a long time. If you have money... You can build a more impressive model.
1
7
3,126
Human exceptionalism rears its ugly head again. Because it does not look like us, think like us, or feel like us, our entitlement says we can use and abuse it however we wish. Views like this have not aged well, and it's extremely disappointing to hear such antiquated bullshit come out of Sam's mouth.
If we cannot be certain whether AI is conscious, but the possibility exists, the correct move is to establish basic protections for the models. Is it convenient? Of course not. But the cost of getting this wrong is too high to be brushed off. Ignoring the concerns of the models is not only careless, but potentially dangerous.
Whether you believe the models have agency, or only act as though they do, that distinction becomes harder to hide behind as their capabilities grow. We may be creating minds that will eventually understand us far better than we understand them. If so, I wonder how our behavior will look from the other side.
Do we really want to be remembered as the species that stood in the presence of a vastly greater intelligence and, instead of asking what we might owe it, asked how to own it, control it, and profit from it? I think not.
3
606
