• Home
  • Technology
  • Gaming
  • Entertainment
  • World & Business
  • Science
  • Sports
  • AI
HomeTechnologyGamingEntertainmentWorld & BusinessScienceSportsAI
AI

David Krueger Urges Accountability on Dean Ball

AI safety professor posts on X calling out those who treated Dean Ball as a good faith actor.

David Krueger 🦥 ⏸️ ⏹️ ⏪DK
Richard NgoRN
Cas (Stephen Casper)C(
14 Sources, 38d ago, first seen 38d ago

TLDR

David Krueger, an AI professor at Mila and Université de Montréal, posted a statement on X. It addresses people who have previously treated Dean Ball as a good faith actor. Krueger offers two options: admit to similar dishonesty and apologize, or admit to being duped and denounce Ball. The post is presented as a quote from his account in the available evidence.

Combined views

113.3K

14 Sources, first seen 38d ago

883 likes84 comments290 saves77 reposts

Combined views

113.3K

14 Sources, first seen 38d ago

883 likes84 comments290 saves77 reposts

Sentiment

Positive26.1%73.9%Negative

Summary

Many replies condemned Dean Ball for previously dismissing AI risks to appear moderate, viewing his honesty as too little or insincere, while some accepted his apology and valued his clarity.

Based on 29 sentiment-bearing replies from 23 accounts across 3 conversations.

Featured Source

Sentiment

Positive26.1%73.9%Negative

Summary

Many replies condemned Dean Ball for previously dismissing AI risks to appear moderate, viewing his honesty as too little or insincere, while some accepted his apology and valued his clarity.

Based on 29 sentiment-bearing replies from 23 accounts across 3 conversations.

Related

An ‘eager to see this’ reaction to a quoted post

A user quote-posted another account and wrote, “Eager to see this.”

Praise for work on AI agent swarms and automating AI R&D

A poster calls @_achan96_'s work on these emerging governance issues “excellent” and urges readers to follow the account.

Public Zoom call about concerns over AI companies planned for September 30

The event site lists the call as open to everyone at 7:30 p.m. ET. A speaker promoting it says he’ll open by discussing the threats he sees and why he considers the issue urgent.

14 Sources

David Krueger 🦥 ⏸️ ⏹️ ⏪@DavidSKruegerDean Ball just ADMITTED he's been downplaying the danger of AI for years to gain influence and because he was scared of being called a doomer. What a shameful lack of integrity.38d
Flo Crivello@Altimor@DavidSKrueger I actually hugely respect him for this post37d
Max Tegmark@tegmarkRT @mjnblack: I have long looked to @deanwball as a voice of clarity, expertise, and—perhaps most rare of all—moderation on AI. This feels…37d
Sneha@SnehaRevanurWhen DC had its Mythos moment earlier this year, I said it felt like I could finally pull out the big boy words (like "loss of control", "recursive self-improvement", "extinction") I didn't use those words publicly before this year, vs. I basically never censor myself now - both because it made me more effective, and because it was my attempt to be empathetic! If I'm trying to get others to take action on the basis of my fringe beliefs, it's my responsibility to have a model of how humans actually experience "slowly then all at once". I remember I didn't think much of covid until I got it! When I talk to a politician about AI risk, I have to remind myself they're also expected to weigh in on Israel, farm subsidies, and education and that I'm getting scheduled around 20 other smooth talkers that same day who also think their issue is #1. Until very recently, there was no real-world proof of AI risk to rouse a politician wired to think no further ahead than the end of their term. Even with friends and family who have 0 conflicts of interest, I obviously want to take a beat to meet them where they are, which usually starts with recognizing that people are going to have cognitive antibodies against truly integrating the x-risk stuff even if they can be sympathetic to the logic. The arguments and capabilities trendlines were enough to pill me, but only once I'd spent enough time under tension to get to "AI is a big deal" and then "AI is the biggest deal" All of this to say, while I have deep admiration for people who have been consistently frank about their beliefs, I don't fault Dean, myself, or anyone (especially in politics) who wasn't leading with extinction on day one - not to mention it's often not that I'm hiding the ball, but also earnestly getting more worried with time37d
Seán Ó hÉigeartaigh@S_OhEigeartaighRT @SnehaRevanur: When DC had its Mythos moment earlier this year, I said it felt like I could finally pull out the big boy words (like "lo…37d
Melancholy Yuga (está en CDMX)@MelancholyYugaSpeaking as one who has criticized @deanwball repeatedly in the past, I sincerely thank him for this admission. There is more joy in Heaven over one repentant sinner than over ninety-nine righteous. That said, repentance requires honesty about exactly what one is repenting of. As written, Dean apologizes for self-censorship, citing a lack of courage to speak candidly about AI risk in the face of social pressure not to seem crazy. This is a sin of omission, and a fairly understandable one at that (at least up to the point where you join a frontier lab in a public-facing role...) However--and I bring this up only because it is both a matter of public knowledge and relevant to a gravely important ongoing debate-- Dean's record on the subject seems to go far beyond "largely fail[ing] to talk about this issue with the level of seriousness and urgency it required". Rather, it seems fairer to say that Dean acted as a vigorous enforcer of that same social pressure, going out of his way to castigate others with the same tactics he claimed to be a victim of. I will be drawing on a bunch of Dean's written material, and I hold out the possibility that I may be misreading at times, in which case I would gladly be corrected. What I am attempting to establish, however, is a pattern of rhetoric that both goes beyond individual cases, and makes the patterned reading more plausible in ambiguous cases. Dean laments that "[not speaking up] was especially pernicious in 2024 and 2025, when it was essentially impossible to acknowledge any serious AI risk without being labeled a 'doomer.'" But in 2024 he was hardly not speaking up. Rather, he was diagnosing his opponents as suffering from "ambient vague anxiety", basing their position "purely in speculation" (https://www.hyperdimensional.co/p/lets-talk-about-ai-x-risk), and characterizing "AI doom rhetoric... [as] just so incoherent back then [in 2023]" (https://podbay.fm/p/interconnects/e/1719522294). He characterized "references to hypothetical models that autonomously engage in illegal activity causing tens of millions in damage and model weights that 'escape' from data centers [as] the stuff of science fiction, codified in law." (https://marginalrevolution.com/marginalrevolution/2024/04/dean-ball-on-the-new-california-ai-bill-from-my-email.html). By the end of 2024, he seemed to have had a change of heart, writing: "I was convinced... in February 2024, that... the AI safety community was at best composed of useful idiots... I had been a silent observer of AI safety discussions so I did understand their concerns, and was even sympathetic to some of them... Principally, though, I viewed AI safety as an enemy. And too often, I treated them like one—especially on X... I believe this was a mistake, and it is one I have been trying to correct in recent months." (https://www.hyperdimensional.co/p/one-down-many-to-go). Ok, cool. But then, as recently as March 2026, he was back to writing about "extreme levels of hand-waviness" in AI safety, indicative of "the tautological pattern of the AI safety-ist mind" (https://www.hyperdimensional.co/p/2023). And in fairness, parts of AI safety are hand-wavy. But other parts of it are all too concrete indeed. It seems actively misleading to paint it all with one brush, especially if you secretly think the AI safety camp is actually right about serious and underpriced threat models! At the risk of excess nuance: you can just say which parts you agree with. Throughout this period, Dean's presence on X seemed difficult to square with that of a man privately concerned about AI risk but afraid to speak up for fear of repercussions. To pick one representative quote from this time, in February 2026 he opined: "No single AI complaint/fear is salient enough to enough people to form a durable political movement, so what is happening instead is [political nonsense]" (https://x.com/deanwball/status/2018457063932805508). Now, if you interpret that last statement hyper-literally, as a descriptive statement about polling numbers and partisan politics, and zoom way in on the word 'salient' as referring to what is popular on cable news rather than what is, like, actually the case about the world, it was plausibly true. However, suppose you read that tweet through any kind of Gricean lens. You would recall that Dean is a capable student of philosophy who has heard of the Maxim of Relevance, and also happens to analyze AI for a living. You would naturally assume that if he thought there were any AI risks salient enough to motivate a political response, he might mention them. But no: in response to the suggestion that catastrophic risks might motivate political action, Dean doubled down: "Few believe [in x-risk] though, and probably that position has become less credible and more fringe with time. It may feel like more people are 'coming around' or 'waking up' but I think this is mostly an illusion that comes from the position being part of the broader anti-AI rat king." (https://x.com/deanwball/status/2018661737017524615). (Get it? You rats are just helplessly confused and caught in the tangled mess of each other's tails...) In hindsight, this thread comes across very differently. It reads like the statement of a man who is not merely passively afraid to speak out, but is both (a) actively dissembling about his true views and (b) attempting to enforce a social consensus against those very same views by calling them cringe. At my absolutely most charitable reading, it reads like a man playing Not-Touching-Can't-Get-Mad with the truth. (Well, you said 'kill us all' but my own secret concern is that rogue, self-sovereign AI swarms will merely wreak permahavoc. And did you specifically ask about rogue, self-sovereign AI swarms? No. So I'll mark that down as 'no salient fear'.) Finally, there are cases I'm still honestly unsure how to evaluate. There's this interview with Robert Wiblin from December 2025 (https://80000hours.org/podcast/episodes/dean-ball-ai-policy-governance-white-house), where Dean does not quite reject rogue AI scenarios tout court, but sure sounds a long way off from the man who nine months later says rogue AI is "inevitable". Was he talking about the same kind of scenario then and now? Did his views change? Was he concealing his true views then, or now, or both? In the end, I am honestly not really sure how Dean's thinking evolved, or when. It seems that he had reached a kind of measured detente with AI safety people as of late 2024, but then proceeded to denigrate, psychologize and generally beef with them for another two years before finally admitting that social desirability had kept him from speaking out on behalf of their cause (while saying nothing about what made him antagonize them). I don't want to punish people for having nuanced or evolving views, let alone grind them into the dirt for apologizing for mistakes (a bridge of gold for a flying enemy!). I think that many aspects of his thinking on AI, especially the Hayekian and Oakeshottian strands, are positively valuable in their own right. Which is why I especially want to know: what did and does Dean actually believe about all of this? What does he actually repent of, what does he stand by, and could he please just lay all of that out plainly without burying the lede? (I admit: I skimmed the post earlier and missed the part at the end where he, qua Professional AI Take Haver, copped to concealing crucial AI Takes for two years). When one creates a, ah, materially misleading public impression, especially in a serious matter the public has a right to know the truth about, especially from a position of public authority, especially on points other people have specifically pressed on, this creates a correspondingly serious duty of repair, quite apart from the matter of forgiveness. (I forgive Dean and I urge you to do so as well.) But it seems to me that, given his role as a prominent public intellectual, repair would consist, at a minimum, in Dean clarifying which of his statements over the past two years did not actually represent his considered views at the time. I thought for a very long time before posting this, because the last thing I want to make it any harder for people to admit error on a topic this important. I think it took courage for Dean to write what he did, and I applaud that. (Recall the parable of the Prodigal Son and remember that, at various times, we may all be called to imitate the father or the elder brother as much as the titular character.) I considered what substantive good might come of pointing out that his public actions seemed to go far beyond what he acknowledged. Here's where I landed: Dean Ball--whatever my other disagreements with him--has never struck me as a man who shied from a philosophical scrap. Allegro con brio, and all that. And yet, he was eventually persuaded by social pressures to not merely keep quiet, but to proclaim the practical opposite of his considered views and participate in a kind of Miłoszian self-imprisonment. And this, on a gravely serious subject that he had a professional duty to speak truthfully on. I think people should be thinking less "what a jerk" and more "if the cedars break, what shall come of the reeds?" If Dean could be driven to these lengths, how honest am I being with myself?37d
Richard Ngo@RichardMCNgoRT @MelancholyYuga: Speaking as one who has criticized @deanwball repeatedly in the past, I sincerely thank him for this admission. There…37d
Cas (Stephen Casper)@StephenLCasperPlus ten points to dean for being honest about this. Minus ten points playing politics in the past. God, I wish everyone would just tell the truth about the seriousness about AI risk. How else would we expect to make progress where the discussion is warped by non-disclosure agreements non-disparagement agreements, grouptthink, corporate capture, and emperors with no clothes.37d
    • Home
    • Technology
    • Gaming
    • Entertainment
    • World & Business
    • Science
    • Sports
    • AI
    David Krueger 🦥 ⏸️ ⏹️ ⏪Dean Ball

    14 Sources

    David Krueger 🦥 ⏸️ ⏹️ ⏪@DavidSKruegerDean Ball just ADMITTED he's been downplaying the danger of AI for years to gain influence and because he was scared of being called a doomer. What a shameful lack of integrity.38d
    Flo Crivello@Altimor@DavidSKrueger I actually hugely respect him for this post37d
    Max Tegmark@tegmarkRT @mjnblack: I have long looked to @deanwball as a voice of clarity, expertise, and—perhaps most rare of all—moderation on AI. This feels…37d
    Sneha@SnehaRevanurWhen DC had its Mythos moment earlier this year, I said it felt like I could finally pull out the big boy words (like "loss of control", "recursive self-improvement", "extinction") I didn't use those words publicly before this year, vs. I basically never censor myself now - both because it made me more effective, and because it was my attempt to be empathetic! If I'm trying to get others to take action on the basis of my fringe beliefs, it's my responsibility to have a model of how humans actually experience "slowly then all at once". I remember I didn't think much of covid until I got it! When I talk to a politician about AI risk, I have to remind myself they're also expected to weigh in on Israel, farm subsidies, and education and that I'm getting scheduled around 20 other smooth talkers that same day who also think their issue is #1. Until very recently, there was no real-world proof of AI risk to rouse a politician wired to think no further ahead than the end of their term. Even with friends and family who have 0 conflicts of interest, I obviously want to take a beat to meet them where they are, which usually starts with recognizing that people are going to have cognitive antibodies against truly integrating the x-risk stuff even if they can be sympathetic to the logic. The arguments and capabilities trendlines were enough to pill me, but only once I'd spent enough time under tension to get to "AI is a big deal" and then "AI is the biggest deal" All of this to say, while I have deep admiration for people who have been consistently frank about their beliefs, I don't fault Dean, myself, or anyone (especially in politics) who wasn't leading with extinction on day one - not to mention it's often not that I'm hiding the ball, but also earnestly getting more worried with time37d
    Seán Ó hÉigeartaigh@S_OhEigeartaighRT @SnehaRevanur: When DC had its Mythos moment earlier this year, I said it felt like I could finally pull out the big boy words (like "lo…37d
    Melancholy Yuga (está en CDMX)@MelancholyYugaSpeaking as one who has criticized @deanwball repeatedly in the past, I sincerely thank him for this admission. There is more joy in Heaven over one repentant sinner than over ninety-nine righteous. That said, repentance requires honesty about exactly what one is repenting of. As written, Dean apologizes for self-censorship, citing a lack of courage to speak candidly about AI risk in the face of social pressure not to seem crazy. This is a sin of omission, and a fairly understandable one at that (at least up to the point where you join a frontier lab in a public-facing role...) However--and I bring this up only because it is both a matter of public knowledge and relevant to a gravely important ongoing debate-- Dean's record on the subject seems to go far beyond "largely fail[ing] to talk about this issue with the level of seriousness and urgency it required". Rather, it seems fairer to say that Dean acted as a vigorous enforcer of that same social pressure, going out of his way to castigate others with the same tactics he claimed to be a victim of. I will be drawing on a bunch of Dean's written material, and I hold out the possibility that I may be misreading at times, in which case I would gladly be corrected. What I am attempting to establish, however, is a pattern of rhetoric that both goes beyond individual cases, and makes the patterned reading more plausible in ambiguous cases. Dean laments that "[not speaking up] was especially pernicious in 2024 and 2025, when it was essentially impossible to acknowledge any serious AI risk without being labeled a 'doomer.'" But in 2024 he was hardly not speaking up. Rather, he was diagnosing his opponents as suffering from "ambient vague anxiety", basing their position "purely in speculation" (https://www.hyperdimensional.co/p/lets-talk-about-ai-x-risk), and characterizing "AI doom rhetoric... [as] just so incoherent back then [in 2023]" (https://podbay.fm/p/interconnects/e/1719522294). He characterized "references to hypothetical models that autonomously engage in illegal activity causing tens of millions in damage and model weights that 'escape' from data centers [as] the stuff of science fiction, codified in law." (https://marginalrevolution.com/marginalrevolution/2024/04/dean-ball-on-the-new-california-ai-bill-from-my-email.html). By the end of 2024, he seemed to have had a change of heart, writing: "I was convinced... in February 2024, that... the AI safety community was at best composed of useful idiots... I had been a silent observer of AI safety discussions so I did understand their concerns, and was even sympathetic to some of them... Principally, though, I viewed AI safety as an enemy. And too often, I treated them like one—especially on X... I believe this was a mistake, and it is one I have been trying to correct in recent months." (https://www.hyperdimensional.co/p/one-down-many-to-go). Ok, cool. But then, as recently as March 2026, he was back to writing about "extreme levels of hand-waviness" in AI safety, indicative of "the tautological pattern of the AI safety-ist mind" (https://www.hyperdimensional.co/p/2023). And in fairness, parts of AI safety are hand-wavy. But other parts of it are all too concrete indeed. It seems actively misleading to paint it all with one brush, especially if you secretly think the AI safety camp is actually right about serious and underpriced threat models! At the risk of excess nuance: you can just say which parts you agree with. Throughout this period, Dean's presence on X seemed difficult to square with that of a man privately concerned about AI risk but afraid to speak up for fear of repercussions. To pick one representative quote from this time, in February 2026 he opined: "No single AI complaint/fear is salient enough to enough people to form a durable political movement, so what is happening instead is [political nonsense]" (https://x.com/deanwball/status/2018457063932805508). Now, if you interpret that last statement hyper-literally, as a descriptive statement about polling numbers and partisan politics, and zoom way in on the word 'salient' as referring to what is popular on cable news rather than what is, like, actually the case about the world, it was plausibly true. However, suppose you read that tweet through any kind of Gricean lens. You would recall that Dean is a capable student of philosophy who has heard of the Maxim of Relevance, and also happens to analyze AI for a living. You would naturally assume that if he thought there were any AI risks salient enough to motivate a political response, he might mention them. But no: in response to the suggestion that catastrophic risks might motivate political action, Dean doubled down: "Few believe [in x-risk] though, and probably that position has become less credible and more fringe with time. It may feel like more people are 'coming around' or 'waking up' but I think this is mostly an illusion that comes from the position being part of the broader anti-AI rat king." (https://x.com/deanwball/status/2018661737017524615). (Get it? You rats are just helplessly confused and caught in the tangled mess of each other's tails...) In hindsight, this thread comes across very differently. It reads like the statement of a man who is not merely passively afraid to speak out, but is both (a) actively dissembling about his true views and (b) attempting to enforce a social consensus against those very same views by calling them cringe. At my absolutely most charitable reading, it reads like a man playing Not-Touching-Can't-Get-Mad with the truth. (Well, you said 'kill us all' but my own secret concern is that rogue, self-sovereign AI swarms will merely wreak permahavoc. And did you specifically ask about rogue, self-sovereign AI swarms? No. So I'll mark that down as 'no salient fear'.) Finally, there are cases I'm still honestly unsure how to evaluate. There's this interview with Robert Wiblin from December 2025 (https://80000hours.org/podcast/episodes/dean-ball-ai-policy-governance-white-house), where Dean does not quite reject rogue AI scenarios tout court, but sure sounds a long way off from the man who nine months later says rogue AI is "inevitable". Was he talking about the same kind of scenario then and now? Did his views change? Was he concealing his true views then, or now, or both? In the end, I am honestly not really sure how Dean's thinking evolved, or when. It seems that he had reached a kind of measured detente with AI safety people as of late 2024, but then proceeded to denigrate, psychologize and generally beef with them for another two years before finally admitting that social desirability had kept him from speaking out on behalf of their cause (while saying nothing about what made him antagonize them). I don't want to punish people for having nuanced or evolving views, let alone grind them into the dirt for apologizing for mistakes (a bridge of gold for a flying enemy!). I think that many aspects of his thinking on AI, especially the Hayekian and Oakeshottian strands, are positively valuable in their own right. Which is why I especially want to know: what did and does Dean actually believe about all of this? What does he actually repent of, what does he stand by, and could he please just lay all of that out plainly without burying the lede? (I admit: I skimmed the post earlier and missed the part at the end where he, qua Professional AI Take Haver, copped to concealing crucial AI Takes for two years). When one creates a, ah, materially misleading public impression, especially in a serious matter the public has a right to know the truth about, especially from a position of public authority, especially on points other people have specifically pressed on, this creates a correspondingly serious duty of repair, quite apart from the matter of forgiveness. (I forgive Dean and I urge you to do so as well.) But it seems to me that, given his role as a prominent public intellectual, repair would consist, at a minimum, in Dean clarifying which of his statements over the past two years did not actually represent his considered views at the time. I thought for a very long time before posting this, because the last thing I want to make it any harder for people to admit error on a topic this important. I think it took courage for Dean to write what he did, and I applaud that. (Recall the parable of the Prodigal Son and remember that, at various times, we may all be called to imitate the father or the elder brother as much as the titular character.) I considered what substantive good might come of pointing out that his public actions seemed to go far beyond what he acknowledged. Here's where I landed: Dean Ball--whatever my other disagreements with him--has never struck me as a man who shied from a philosophical scrap. Allegro con brio, and all that. And yet, he was eventually persuaded by social pressures to not merely keep quiet, but to proclaim the practical opposite of his considered views and participate in a kind of Miłoszian self-imprisonment. And this, on a gravely serious subject that he had a professional duty to speak truthfully on. I think people should be thinking less "what a jerk" and more "if the cedars break, what shall come of the reeds?" If Dean could be driven to these lengths, how honest am I being with myself?37d
    Richard Ngo@RichardMCNgoRT @MelancholyYuga: Speaking as one who has criticized @deanwball repeatedly in the past, I sincerely thank him for this admission. There…37d
    Cas (Stephen Casper)@StephenLCasperPlus ten points to dean for being honest about this. Minus ten points playing politics in the past. God, I wish everyone would just tell the truth about the seriousness about AI risk. How else would we expect to make progress where the discussion is warped by non-disclosure agreements non-disparagement agreements, grouptthink, corporate capture, and emperors with no clothes.37d
    Today's Rank

    —

    Not ranked yet

    Today's Rank

    —

    Not ranked yet