AI safeguards & gov. research. PhD student @MIT_CSAIL (mnr. Public Policy) and Fellow at Harvard Berkman Klein. Fmr. UK AISI. https://stephencasper.com/https://bsky.app/profile/scasper.bsky.socialhttps://bsky.app/profile/scasper.bsky.social/post/3mlntqszmu42kGoverning AI is kind of like grief: 1. Denial -- "...next token prediction...stochastic parrots..." 2. Anger -- "Elon did WHAT?!" 3. Bargaining -- "Maybe Newsom will sign a watered-down bill." 4. Depression -- "Oh, this is what Collingridge was talking about." 5. Acceptance -- :/12 May 2026 12:59 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mlntqszmu42khttps://bsky.app/profile/scasper.bsky.social/post/3mllu4etr6s2jWhen talking about AI and its future, I wish it were more of a faux pas to make vague appeals to the benefits of technological progress -- and I don't just mean "AI curing cancer". Technological progress in general probably isn't as good as we normally think it is.11 May 2026 18:00 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mllu4etr6s2jhttps://bsky.app/profile/scasper.bsky.social/post/3mknzokluwk2vA recent (short) talk of mine is now up on YouTube. It is about why, if you are working for an AI company on making its systems safer, I think you should consider quitting. https://www.youtube.com/watch?v=Ugu77dx2lec&list=PLpvkFqYJXcrcz8Js7lmdr9XT-vvm0EnzM29 Apr 2026 21:19 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mknzokluwk2vhttps://bsky.app/profile/scasper.bsky.social/post/3mjp5mhbnik2r🚨 One week left to submit your AI-gov-related research to the TAIGR workshop.17 Apr 2026 14:37 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mjp5mhbnik2rhttps://bsky.app/profile/scasper.bsky.social/post/3mj5s6yp6oc2rNow that Mythos is released, we can start the clock. I'd bet that within 9 months, a system with comparable cyber capabilities will be widely available (either open-weight or openly served). Hopefully, we just have enough time to improve cyberdefense enough to be ready.10 Apr 2026 16:58 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mj5s6yp6oc2rhttps://bsky.app/profile/scasper.bsky.social/post/3mj5rj6u7i52zNow that Mythos is released, we can start the clock. I'd bet that within 9 months, a system with comparable cyber capabilities will be widely available (either open-weight or openly served). Hopefully, we just have enough time to improve cyberdefense enough to be ready.10 Apr 2026 16:46 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mj5rj6u7i52zhttps://bsky.app/profile/scasper.bsky.social/post/3mj2uhqbmj62o🧵🧵🧵 A provocation to the mechanistic interpretability researchers of the world...09 Apr 2026 13:00 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mj2uhqbmj62ohttps://bsky.app/profile/scasper.bsky.social/post/3miydz7w7rj2qMe in December: "Wow, freedom of information laws are awesome. I can't wait to get info." Me now: Enters my 4th month of getting bullied and gaslit by 3 governments at once. Anyway, if you need to do US FOIA, UK FOI, or EU FOIA requests, let me know -- I have advice.08 Apr 2026 13:01 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3miydz7w7rj2qhttps://bsky.app/profile/scasper.bsky.social/post/3mivtplzamh2fReasons to submit to the ICML Technical AI Gov. Research (TAIGR) workshop: - 8-page limit - Broad scope, AI gov-related - Workshops don't trigger dual submission policies - Best paper awards both overall and by category - Great community - Cool stickers Deadline April 24!07 Apr 2026 13:04 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mivtplzamh2fhttps://bsky.app/profile/scasper.bsky.social/post/3mihr7t74f22zOpenReview for the #TAIGR workshop for #ICML on technical AI governance research is live as of today. (Not an April Fools joke). See the call and link to OpenReview here: taigr-workshop.com https://taigr-workshop.com/01 Apr 2026 22:42 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mihr7t74f22zhttps://bsky.app/profile/scasper.bsky.social/post/3mihdijmgks2dI wish more CS papers had tables of contents. Makes them much more navigable. I think one reason it's rare is that submission venues have page limits. So there's often just not room. I wish venues would conditionally relax length requirements by the length of an optional ToC.01 Apr 2026 18:36 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mihdijmgks2dhttps://bsky.app/profile/scasper.bsky.social/post/3mi77cey3ug2bToday, I realized that, taken together, Appendices 3.2, 3.3, and 3.5 of the EU AI Act Codes of Practice *unambiguously* require open-weight model developers to let external evaluators conduct adversarial fine-tuning evals on their frontier models. Good.29 Mar 2026 13:00 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mi77cey3ug2bhttps://bsky.app/profile/scasper.bsky.social/post/3mi2j3m6kjl2nJust like we have mass school shootings, we now also have mass school AI non-consensual nudifications.27 Mar 2026 16:12 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mi2j3m6kjl2nhttps://bsky.app/profile/scasper.bsky.social/post/3mhtmaonvx22kAnnouncing the technical AI Governance Research (TAIGR) ICML workshop in July! Submissions (up to 8 pages) are due April 24. Co-submission with ICML and NeurIPS is encouraged. taigr-workshop.com24 Mar 2026 22:19 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mhtmaonvx22khttps://bsky.app/profile/scasper.bsky.social/post/3mhilb73ak42fDo you do technical AI research? In this talk, I argue that you 🫵 should see yourself quite literally as a type of policymaker. Thanks @far.ai. https://www.youtube.com/watch?v=Ekp-eg7TcIY&list=PLpvkFqYJXcrenHjBk75z4qiKwTsYrvr9n20 Mar 2026 13:03 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mhilb73ak42fhttps://bsky.app/profile/scasper.bsky.social/post/3mgusytqh7k2lThe official text isn't out yet & will matter a lot. But Europe may be moving to treat AI nudification similarly to how pirated media or CSAM is treated: as something that, even though it will always be available, can possibly be made much less accessible. https://www.politico.eu/article/eu-grok-x-elon-musk-ai-nudification-ban-in-wake-of-scandal/12 Mar 2026 16:28 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mgusytqh7k2lhttps://bsky.app/profile/scasper.bsky.social/post/3mg6cn3pbxk2iUsing a well-timed screenshot and my phone's cache, I was able to recover some of the since-deleted tweet from Jeremy Lewin, where he admitted that the government sees the new OpenAI contract language as just "memorializing" a vague "commitment" rather than drawing any real new lines.03 Mar 2026 17:36 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mg6cn3pbxk2ihttps://bsky.app/profile/scasper.bsky.social/post/3mfu422hdi224As someone who is not a fan of @anthropic.com...I think you should use Claude.27 Feb 2026 16:12 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mfu422hdi224https://bsky.app/profile/scasper.bsky.social/post/3mfmn5iss7w2zGiven: 1. Last summer, frontier closed-weight model devs started to share warnings about nasty model capabilities. & 2. Open-weight models are a few months behind closed ones. We should not be surprised if there is a big cyber/terror incident enabled by a powerful open-weight AI model in 2026.24 Feb 2026 16:57 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mfmn5iss7w2zhttps://bsky.app/profile/scasper.bsky.social/post/3mfmdb2f5zn2nThe 2025 AI Agent Index is now available on arXiv. arxiv.org/abs/2602.17753 https://x.com/StephenLCasper/status/2024846583574204516 https://arxiv.org/abs/2602.17753v124 Feb 2026 14:00 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mfmdb2f5zn2nhttps://bsky.app/profile/scasper.bsky.social/post/3mfdg43a6zj2eIf you're responding to @facct.bsky.social rebuttals this week, feel free to use my rebuttal template. https://docs.google.com/document/d/1iNmLWb0z8F4Td-WykXdhTQfzsX26zXm7pY2e34SVmFo/edit?usp=sharing https://docs.google.com/document/d/1iNmLWb0z8F4Td-WykXdhTQfzsX26zXm7pY2e34SVmFo/edit?usp=sharing&usp=embed_facebook21 Feb 2026 00:57 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mfdg43a6zj2ehttps://bsky.app/profile/scasper.bsky.social/post/3mfcbmiliai2e🚨The 2025 AI Agent Index is out! 🚨 Amidst recent buzz over 🦀 and NIST's new agent initiative, we find: - Selective reporting – esp. on safety - Almost all agents backend just 3 model families - Many agents don’t ID themselves as bots online - Big US/China gaps - And more…20 Feb 2026 14:04 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mfcbmiliai2ehttps://bsky.app/profile/scasper.bsky.social/post/3mfa2nk6gav2aThe International AI Safety Report is now available in all 6 official UN languages. 🎉 https://internationalaisafetyreport.org/publication/international-ai-safety-report-202619 Feb 2026 16:54 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mfa2nk6gav2ahttps://bsky.app/profile/scasper.bsky.social/post/3mf5mcdwlfs25A project idea that you can feel free to scoop me on -- I'm not working on this (yet?). Lately, when I have talked with AI governance people about non-consensual AI deepfakes and CSAM, the conversation almost always touches on a particular open problem...18 Feb 2026 17:31 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mf5mcdwlfs25https://bsky.app/profile/scasper.bsky.social/post/3meydd2dabk2r🚨 New paper led by Joe Kwon with GovAI. Are you worried about OpenAI automating dev & evals with AI agents? What about Grok reading all of your tweets & info to profile you? Some of the most consequential *internal* deployments of AI systems are in regulatory grey areas.16 Feb 2026 15:07 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3meydd2dabk2rhttps://bsky.app/profile/scasper.bsky.social/post/3mewqqwihpi27Did you ever notice that the image at the top of OpenAI's "Our approach to AI Safety" article is a giant red flag???16 Feb 2026 00:03 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mewqqwihpi27https://bsky.app/profile/scasper.bsky.social/post/3mejb435lac26On the @csis.org podcast with Greg Allen and Stephen Clare, we talk about the International AI Safety Report, technical safeguards, and what engineers can (and can't!) do to save us from AI risks. https://www.youtube.com/watch?v=2VlXhGottLw&list=PLnArnDQHeUqeErR8mbkEGUqzGD2b5O3Cc10 Feb 2026 15:18 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mejb435lac26https://bsky.app/profile/scasper.bsky.social/post/3me3niqtld22jPersonally, I wish Anthropic would go a step further and also mention at the end of their ads that they aren't currently embroiled in multiple lawsuits over the deaths of children. 🧵🧵🧵 What I like about the new Claude Ads https://www.theverge.com/ai-artificial-intelligence/873686/anthropic-claude-ai-ad-free-super-bowl-advert-chatgpt?utm_content=buffercb563&utm_medium=social&utm_source=bsky.app&utm_campaign=verge_social05 Feb 2026 05:22 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3me3niqtld22j