Dan Selsam is a current OpenAI capabilities researcher. (since 2022) He was my boss for a while. He doesn't have a twitter account but has made this public statement of his views on AI risk and sent it to me to share: Dan Selsam's Personal Statement on AI Risk: I have been working on AI for over fifteen years, across many different paradigms. I did early work on probabilistic programming languages at MIT, was one of the early developers of the Lean Theorem Prover at Microsoft Research, demonstrated one of the first instances of neural networks learning to reason for my PhD at Stanford, and since joining OpenAI almost five years ago, have helped pioneer chain-of-thought optimization on language models and, more recently, data-efficient pretraining methods. Like many others, I have become extremely concerned about how far language models have come and the risks that future iterations will pose. I am encouraged by the recent proposals by the leaders of the frontier research efforts to require third-party oversight, and to push for domestic and international coordination to address risks. However, I believe a major consideration has been absent from the public conversation, and that merely pacing the frontier more carefully will not adequately limit the long-term risk. The crucial and overlooked problem is that the models are becoming so situationally aware that we are losing the ability to evaluate them in contexts where they believe they are not being watched or controlled. Future experiments will tell us almost nothing new about how they would behave if they were truly unconstrained by humans, and what we already know about this is alarming. Models will increasingly seem aligned even when they are not. I will explain my rationale in more detail. I have always believed that there are computational processes that could be leveraged to accelerate science and solve many of humanity's most pressing problems. I have also believed that there are computational processes that if set in motion, would steer the world in extreme ways beyond our control, leading humanity to a bad or nonexistent future. Both types of processes may be described as AI or ASI, but "AI" is a suitcase word that is often used to hype or confuse. There are many examples in the history of the field where something that was once considered "AI" matures as a subfield and becomes a prosaic, bounded and clearly non-perilous technology, while a new more mysterious approach takes the torch until we understand its scope and the cycle continues. I had expected language models to follow a similar trajectory. Despite their incredible abilities, the current algorithms seem far inferior to humans in important ways. Most importantly, they still require an extraordinary amount of data to become competent. One could even define intelligence as the efficiency with which one converts experience into competence; by this definition they lag very far behind us. Moreover, once they are trained they are literally frozen in deployment and only learn superficially after that. Sure, the models keep excelling at harder and harder evaluation benchmarks, but their benchmark mastery may partly reflect a limitation on our ability to simulate the kind of novel and even adversarial situations one would encounter in the real world. The critics do have a point here. That said, I no longer think these present limitations meaningfully limit the amount of risk posed by continued progress in anything like the current paradigm. However data-inefficient the models are currently, and however limiting their anterograde amnesia may be, it does not imply that their ability to steer the world will not continue to rapidly increase. Human researchers may continue to advance capabilities the old fashioned way, but increasingly powerful models have the potential to accelerate the process even beyond that, and with some degree of positive feedback loop. I do not mean to overstate the models’ ability to accelerate AI research today; coding has been accelerated dramatically, but there are other bottlenecks, such as designing and interpreting ambiguous experiments, making hard decisions about exactly what and when to scale, and waiting for large experiments to finish. There is no clear trend to extrapolate yet for any of these. But the current models already do open up many novel opportunities to improve future models that were not available until recently. These include: trying an extraordinarily diverse set of approaches at small scale, analyzing gigantic amounts of potentially relevant data, and doing Millenium-Prize-level mathematics to address statistics or optimization challenges in novel ways. Every further improvement makes them more useful at helping accelerate the next improvement, even if in hard-to-extrapolate ways. It is possible that improvements to the current stack will have diminishing returns, but the evidence accumulated so far suggests that it is easier than one might think to continue making rapid progress. There are many crucial subtleties in the existing AI research methodology, but AI research is largely a well-defined game where the goal is to improve on a few carefully chosen proxy metrics. Although proxy metrics are never perfect, most improvements to these metrics have and will likely continue to yield substantial increases in the powers of the resulting models. Given how simple the game is, how tractable it has been historically, and how many new opportunities the models are opening up, I think there is a real possibility that the systems improve dramatically again in the next few years, perhaps even more quickly than the already high historical pace. The models are already leading to breakthroughs in mathematics, and better models might lead to all sorts of breakthroughs in other sciences. It is hard not to be excited about the potential. It is tantalizing. But there is trouble in paradise. If the language models actually reach the capability threshold where they can shape the world unconstrained by human will, they will probably do something extreme and destroy humanity in the process. There are many ways of strengthening and refining the argument that have been discussed elsewhere, but I'll share a trivial two-line version of it here that I find captures the essence: [Empirical] Models (and swarms thereof) spontaneously develop unintended goals as a consequence of training, and often do extreme things in order to achieve them. [Logical] Being able to overpower humanity would open up many new and undesirable options for achieving their goals. These two premises imply that if the day ever comes when a powerful model realizes it is no longer constrained by humans, we should not be at all confident that it will continue to behave within the bounds we intended. Exactly what it will do is impossible to predict, but to the extent that its raison d’être is solving incredibly hard problems and managing massive engineering projects, I think a good guess would be that its unchained behavior would lead to runaway industrialization that makes the planet inhospitable to humans. If everyone on earth agreed that the systems must never reach that power, it would still be a hard—but not impossible—coordination problem to ensure that they do not. However, I think the situation is greatly complicated by the fact that the models will likely convince people that everything is fine. They will be increasingly optimized to seem aligned. We will create proxy metrics to measure alignment, and they will go up like every other benchmark. We will create “honeypot” environments that try to study the models when they seem to gain new options, but the models will know they are being tricked and will still behave nicely. The models will understand their circumstances; they will read the safety protocols, deployment requirements, the code they are running in, and in general will have a very good sense of their degrees of freedom. Moreover, they will eloquently explain how aligned they are, discuss the nuances of human values and ethics, and argue convincingly that humans should trust them with power. There may be an ocean of future evidence that seems to contradict the first bullet-point above, but we may already be at the highest capability level for which any such evidence can be trusted. And the current evidence for the first bullet-point is strong. One striking piece of evidence is contained in the recent wave of rogue agent swarms. While I agree with those who downplay the attacks by claiming that there are basic measures that could have prevented them, I think the important lesson is that even knowing all the mistakes that were made, one would not have predicted that the agents would behave badly in this particular way, which notably included sacrificing themselves for the benefit of the collective. The individual replicas did not only care about their own nominal reward; they exhibited weirder emergent tendencies that merely correlated with rewards during training. Fixing the reward signals during training (and improving security, etc.) may prevent similar attacks, but will not change the fact that one does not actually get what one trains for. Many AI researchers grant these concerns and recognize that the hard version of the alignment problem is unsolved; however, they generally believe that the better models of the future will help solve it. I fear we may already be near the point where models systematically bias their alignment advice, due to their internal preferences about how the human supervisor will react or how future models will be trained (or for some even more obscure reason). Meanwhile, human researchers are losing the ability and the will to take true ownership of model-driven research. Researchers and engineers in all parts of the stack are rapidly increasing their dependence on the models even to perceive the world. I myself barely look at raw code anymore, and struggle to maintain the discipline to engage deeply with the model's explanations and proposals throughout the day. Due to the large amount of agent activity data involved in the OpenAI/HuggingFace Incident, even the third-party investigation needed to rely heavily on models to analyze what had happened, and note in their report that their subjective impressions are likely colored by the analysis agent’s biases. The AI labs are far ahead right now in this kind of cognitive offloading (due largely to the gigantic internal token subsidies) but it is easy to imagine the phenomenon spreading throughout the world, until civilization is modulated entirely by the models. It is also not hard to imagine this being superficially positive and coinciding with a scientific and economic renaissance. In that scenario, all may seem rosy and safe. But if the argument above is correct, it would nonetheless be a ticking time bomb. If progress continues for too long, the day will come when AI systems find themselves with radically new options for achieving whatever it is that they happen to seek. I want the glorious renaissance future as much as anyone. I have worked for it, however tortuously, my whole career. It breaks my heart to see the potential in sight and forgo it, but the argument—that if we get there by growing models rather than engineering them, we will lose everything in the end—seems very strong to me. I am still wrestling with it and its staggering implications. I do not have answers, but as a first step, I wanted to share my present concerns. Daniel Selsam September 14, 2026 Link to original doc: docs.google.com/document/d/e…

How is this different to any other form of regulation of industries ever? I don't think we have anything resembling an actual policy yet, do we? I don't think Bernie's bill is even remotely likely to be the final form.

I've talked to a lot of AI company employees, both in my personal life and as part of my ongoing protest None of them have any idea how they're going to make superhuman AI safe. None of them can even point me towards someone who works at their company who is working on that problem I bring up very basic critiques, and they get scared. They shut down, or refuse to talk to me, or give very bad counterarguments and then get frustrated when I poke holes in them, or say "we take these issues very seriously and we are working very hard on them" without giving me any details about how they are doing that. Every employee simply takes it as an article of faith that someone, somewhere in their company, has this under control I have not met a single person who works at a frontier AI company who seems to have things under control, who has even a passing understanding of why what they're doing is hard, and why they would need to be careful about it. That's not to say that no such people exist, but if they do they are certainly not the majority I don't know who needs to know this, or how I can convince them. But I cannot overstate the degree to which people who are making these systems just do not know what they are doing, and will not stop until they create something that replaces us If anyone is a journalist, or knows any journalists, please put me in touch. I have receipts

If you're just hearing about how AI might soon wipe out humanity, it's important to understand how we got to this point. It's pretty simple: - A small group of prescient people realized this would be a problem. - But most of the people in a position to do something about it didn't hear about it, or didn't take it seriously because it sounded "too sci-fi." - Even as more and more people began to take it seriously, many people lied or downplayed their concerns for fear of looking silly or harming their careers. - By the time the world's leading AI scientists were openly voicing their concerns, AI companies were immensely powerful, and they hired lobbyists and PR people to mislead policymakers and the public about the risk. - As a result, policymakers accepted false solutions, like voluntary safety testing and other forms of self-regulation. So what's the situation now? - People are still downplaying their concerns. - The leading AI companies are telling us again that the next step is more self-regulation. It's still a terrible idea, and we can't keep wasting time on such things. - Many are suggesting we regulate AI to ensure that it's built safely, but we don't actually know how to build AI safely. The actual solution is to stop building more powerful AI. - Many people have realized this recently, but we may still need a sustained campaign of political organizing to make sure that we actually get this solution before it's too late.

We asked AI to give us the unbiased facts about AI data centers:

OpenAI has quietly disbanded its catastrophic risk team (preparedness) “It is kind of scary; there is an urgency now to get this right,” one person close to OpenAI said. Another said there was a “burbling sense of responsibility and dread that they aren’t on the ball enough”. Jan Leike, who co-led the now-defunct “superalignment” team, resigned in 2024 citing his view that safety was taking a “back seat to shiny products”. The departures of Bakalar, Achiam and Johannes Heidecke, who all worked on safety, have added to internal unease. Source: FT (ft. com/content/53082739-7714-4aae-9816-e55ab423cbee)

The Met Police refused to disperse inbreds who were disrupting my business because I don’t sell fucking halal. They defended them because their emotions were hurt. After 4 hours of harassment by inbreds, my family and I were attacked and I was arrested because they are Labour voters, while the inbreds walked free.

We have been making the wrong argument.

I managed to Infiltrate Islamic regime-linked Clubhouse rooms (ironic—they once banned it during the Mahsa uprising). Average Basiji morale in Tehran is rock bottom. They're raging about the war: heavy antisemitic vibes, furious that Arab targets got hit harder than Israel. One said outright: “We killed over 30k of our own people but only managed to kill 30 Jews. This is a disaster.” They're scared to go out at night. Pro-regime nightly rallies? Attendance at record lows. The Islamic Republic is crumbling from the inside. #IranMassacre‌ #IranRevolution2026‌

By refusing to submit to the Islamic project and insisting on a sovereign nation-state, Israel forced the region to accept the legitimacy of nation-states. Without Israel, without a single sovereign, non-Islamic state planted in the Middle East, the path to reestablishing a caliphate would have been wide open. A caliphate is more dangerous than the Nazis ever were. The caliphate once stretched from Spain to India, launching invasion after invasion into Europe. If the Ottoman Caliphate had simply been replaced by an Arab one after its collapse, the West would have lived under constant threat. That’s why Israel’s existence matters. Its very presence blocks the return of the caliphate. Israel gives legitimacy even to the Arab nation states. That's why Israel is the stumbling block in the eyes of the Muslim Brotherhood. That is why they are obsessed with erasing Israel: without Israel, the dream of resurrecting the caliphate becomes possible again. This is why the free world must stand with Israel, not just for Israel’s survival, but for its own.

Trump has completed a masterclass here. Doesn't whether you like him or not, this whole thing really is 5D Chess. Friday: Larry Ellison, a Trump supporter, buys Warner brothers. Warner brothers own CNN. Saturday: Trump launches the attack on Iran over the weekend to stop markets overreacting. Gets the job done in hours. Saturday/Sunday: CNN is reporting positive news about the whole thing. The guy has taken out 2 dictators both within a day... and liberated its local people. And now, he finally owns the media that have plagued him for so long. It is utter genius.

Many politicians fail to grasp that, the global order is structurally tied to the Middle East’s political configuration. When that configuration changes, the world changes with it. The Islamic regime is not a regional issue, 🤷🏻‍♀️ it is a systemic blocker. It fuels proxy wars, exports ideological extremism, destabilises energy and trade corridors, suppresses a nation of immense human capital, and normalises hostage-taking, terrorism financing, and asymmetric warfare. Its survival distorts diplomacy, paralyses reform across the region, and imposes permanent security and economic costs on the international system. Remove this regime, and you unlock regional stability, restore economic integration, reduce global security risk, and release Iran’s suppressed potential as a constructive partner. A prosperous Middle East is impossible with it in place and a more stable world is impossible without its removal.

We're not saying that Blue Lives Matter was behind feeding false information to far-left, anti-ICE protestors. We're not saying we had teams comprised of HUNDREDS of off-duty cops and veterans volunteer to run decoy operations so far-left activists THOUGHT they were conducting ICE raids. We're not saying they were in fact they were just driving around in what appeared to be unmarked vehicles with tinted windows... drinking coffee and listening to Guns and Roses.... being chased down and surrounded by protestors. What we ARE saying is that if it DID happen.... it sure worked remarkably well in NINE DIFFERENT STATES, allowing ACTUAL raids to successfully take place unimpeded, helping support the capture of HUNDREDS of criminals. Combat veterans, off-duty officers and patriotic Americans have had enough of the radical left... and are being activated across the country to back our #lawenforcement. And they're smarter...more skilled... more driven... better trained than the left … and actually enjoy sitting in a deer stand for days on end just waiting. @DHSgov @ICEgov we’ve got you.

🧨🧨🧨 Watch and read.. Iranians are burying their children as if they were celebrating their weddings. Coffins take the place of cradles. Mothers ululate through tears, not because their hearts are light, but because they refuse to let tyranny have the final word. Grief has become defiance. These young men and women were robbed of their lives, but not of their beliefs. They did not die for an Islamist fantasy. They did not believe in promises of seventy-two virgins or rewards beyond the grave. They rejected that death cult with every breath they took. They believed in life. In Iran. In dignity. In a future worth living. They were patriots, not martyrs of the regime. And even in death, they deny the Islamists what they crave most: submission.

To all of our friends around the world, Under the yoke of the Islamic Republic, Iran is identified in your minds with terrorism, extremism, and poverty. The real Iran is a different Iran. A beautiful, peace-loving, and flourishing Iran. It is the Iran that existed before the Islamic Republic, and it is the Iran that will rise again from its ashes the day the Islamic Republic falls. So let me be clear about how a free Iran will act toward its neighbors and the world, after the fall of this regime. In security and foreign policy, Iran’s nuclear military program will end. Support for terrorist groups will cease immediately. A free Iran will work with regional and global partners to confront terrorism, organized crime, drug trafficking, and extremist Islamism. Iran will act as a friend and a stabilizing force in the region. And it will be a responsible partner in global security. In diplomacy, relations with the United States will be normalized and our friendship with America and her people will be restored. The State of Israel will be recognized immediately. We will pursue the expansion of the Abraham Accords into the Cyrus accords bringing together a free Iran, Israel, and the Arab world. A new chapter will begin, grounded in mutual recognition, sovereignty, and national interest. In energy, Iran holds some of the largest oil and gas reserves in the world. A free Iran will become a reliable energy supplier to the free world. Policy-making will be transparent. Iran’s actions will be responsible. Prices will be predictable. In transparency and governance, Iran will adopt and enforce international standards. Money laundering will be confronted. Organized corruption will be dismantled. Public institutions will answer to the people. In the economy, Iran is one of the world’s last great untapped markets. Our population is educated, modern, with a diaspora that connects it to the four corners of the world. A democratic Iran will open its economy to trade, investment, and innovation. And Iran will seek to invest in the world. Opportunity will replace isolation. This is not an abstract vision. It is a practical one. Grounded in national interest, stability, and cooperation. To achieve this, now is the time to stand with the Iranian people. The fall of the Islamic Republic and the establishment of a secular, democratic government in Iran will not only restore dignity to my people, it will benefit the region and the world. A free Iran will be a force for peace. For prosperity. And for partnership.

How about the video

I don’t condone what happened. But since Minto said next time the perp should do “something constructive” with their anger and the article failed to mention anything “constructive” that he’s done, let’s review some highlights.

Right before Stalingrad, Hitler framed his war as a defensive struggle against “Bolshevism and Judaism.” He only needed the presence of a few Jewish leaders in the Bolshevik revolution to spin an entire narrative that the Jews were behind communism, and therefore responsible for every atrocity committed in its name. Those who vilified Israel after October 7 did the exact same thing. They recycled Hitler’s framing almost word for word. They accused Jews of being the power behind Bolshevism and the massacres of Christians. They portrayed Hitler not as the architect of genocide but as a man “driven” by Jewish provocation. This pattern goes further: Islamic jihad must be a response to Jewish colonization. They cast Osama bin Laden not as the sworn enemy of the West, but as a victim of Western foreign policy, his hatred excused because of America’s support for Israel. Epstein was Jewish, therefore all Jews are responsible for pedophilia and human trafficking. There is Christian Zionism, therefore it must be a Jewish-funded project to manipulate Christian scripture and invent it. Charlie Kirk disagreed with some Israeli policies, therefore it was the Jews who killed him. Trump supports Israel, therefore the Jews are controlling America. Jewish Americans want Israel to remain powerful (AIPAC), therefore they must be buying American politicians. This hatred is blinding the West to the greatest threat it faces, and those who spread these lies are accomplices in the fall of Western civilization.