Computer scientist working on AI safeguards, incidents, & governance research. Assistant professor @harvardkennedy.bsky.social @harvard.edu. https://stephencasper.com/https://bsky.app/profile/scasper.bsky.socialhttps://bsky.app/profile/scasper.bsky.social/post/3mvios7zlkf2a🧵🧵🧵 What if we are on the verge of a "Cyber Cambrian"? (Plus personal updates on my thinking about risk.)14 Sep 2026 17:59 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mvios7zlkf2ahttps://bsky.app/profile/scasper.bsky.social/post/3muhlfbtrwx24After the HF hacking incident and subsequent reporting on it. It seems time to decisively reject the 'AI as a normal technology' hypothesis. Any version of it viable today would be contorted either into a truism or beyond any reasonable interpretation of the word "normal".01 Sep 2026 14:00 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3muhlfbtrwx24https://bsky.app/profile/scasper.bsky.social/post/3mufhicfavo2yWhen I first started grad school, I skimmed all 50-70 new CS arXiv titles daily. Today, there are >300, plus a deluge of news about policy, lawsuits, & other stuff. So I vibe-coded this openly available website to give me a personalized daily digest. https://thestephencasper.github.io/whatiscasreading/31 Aug 2026 17:45 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mufhicfavo2yhttps://bsky.app/profile/scasper.bsky.social/post/3mu6anaydvs2lYesterday, Robert Wright and I discussed what will happen when AI systems become an invasive, parasitic species in cyberspace that undergo digital and cultural evolution.28 Aug 2026 20:53 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mu6anaydvs2lhttps://bsky.app/profile/scasper.bsky.social/post/3mtzvt5nc322fI concur with everyone else that the new reports from OpenAI and METR are bonkers. But I think the most important takeaway from these reports is that OpenAI seems to be so obviously backburning. Predictably but conveniently, these reports were exclusively about the factual sequence of events... https://www.alabamaag.gov/wp-content/uploads/2026/08/OpenAI-Subpoena_Final.pdf27 Aug 2026 03:29 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mtzvt5nc322fhttps://bsky.app/profile/scasper.bsky.social/post/3msvxj4jhjo2x🧵🧵🧵I think that technical safety for closed-weight AI systems is, at this point in time, a solved problem. When I say this, some people look at me like I'm crazy, so let me explain here what I mean by "safety", what the SOTA currently is, and what 6 things can do the trick.12 Aug 2026 20:24 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3msvxj4jhjo2xhttps://bsky.app/profile/scasper.bsky.social/post/3mstqnqt4e227I think this blueprint is exceptionally good. There are a few details I wish it would have added, but this is my favorite policy proposal (if you could call it that) to date related to AI standards/assurance. https://x.com/americans4ri/status/208692183251539152511 Aug 2026 23:16 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mstqnqt4e227https://bsky.app/profile/scasper.bsky.social/post/3mssuy5qh5j2oThis 0% AI-generated paper is now up on arXiv 10 months after we submitted and 5 months after it was accepted to TMLR. Those 10 months of hold, review, and appeal were ultimately pointless. I'm thankful it's up, but I wish that arXiv weren't so broken. arxiv.org/abs/2608.07514 https://arxiv.org/abs/2608.07514v111 Aug 2026 15:00 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mssuy5qh5j2ohttps://bsky.app/profile/scasper.bsky.social/post/3mssob6ba5y2tIf AI companies are serious about pauses, then they should put out documents proposing a 'pause framework' in which they describe what they would do to slow down, how they would make these actions verifiable to third parties, and under what conditions they would do it.11 Aug 2026 13:00 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mssob6ba5y2thttps://bsky.app/profile/scasper.bsky.social/post/3msqipaf27c2vGlad to see this from @sanders.senate.gov. https://www.sanders.senate.gov/wp-content/uploads/AI-Pause-Letter-FINAL.pdf10 Aug 2026 16:15 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3msqipaf27c2vhttps://bsky.app/profile/scasper.bsky.social/post/3msq5ryqcd424I realized something this week: there does not exist a proposed or enacted federal or state frontier AI law in the US that would impose any criminal penalties for an AI company willfully making false or misleading statements to the public about an imminent catastrophic risk.10 Aug 2026 13:00 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3msq5ryqcd424https://bsky.app/profile/scasper.bsky.social/post/3msopkkbbow2wWe are probably just a few months away from some types of cyber-capable AI agents literally becoming a type of parasitic invasive genus in cyberspace that undergo digital and cultural evolution. Biologists and linguists should prepare to study some really crazy stuff.09 Aug 2026 23:13 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3msopkkbbow2whttps://bsky.app/profile/scasper.bsky.social/post/3msdlnxs4m62lNone of the 19 incidents that UK AISI found were from 'helpful-only' models. There is a case to be made that setting Mythos- or GPT 5.6-Sol-level AI cyberagents to run without a robust real-time monitoring setup is an inherently (perhaps abnormally) dangerous activity.05 Aug 2026 13:04 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3msdlnxs4m62lhttps://bsky.app/profile/scasper.bsky.social/post/3msdhvlfhks25“The agents went rogue.” Kinda, but imagine that a zoo had a habit of not shutting the door on animal enclosures or putting elephants behind chicken wire — all with no zookeepers in sight. The fault is not in our stars.05 Aug 2026 11:56 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3msdhvlfhks25https://bsky.app/profile/scasper.bsky.social/post/3mrpuh4begl22🧵 If trends hold, expect a Mythos-level open-weight model around Christmas. Meanwhile, open models are important but also a massive hole in most agendas for safe AI. Here's a 🧵 of my thoughts & research agenda on the technical & political challenges we need to address.28 Jul 2026 16:48 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mrpuh4begl22https://bsky.app/profile/scasper.bsky.social/post/3mrhxyclkrx2yI was wondering if there was any research about how AI companies sometimes perversely keep their safety research to themselves to build a moat around it and gain a competitive edge over competitors. I found one. It's a cool paper. Sharing here in case anyone's interested.25 Jul 2026 13:30 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mrhxyclkrx2yhttps://bsky.app/profile/scasper.bsky.social/post/3mrfftvvbtw24🧵 In AI, we are used to seeing graphs that start to exhibit hockey stick behavior around 2023-2025. But that's a little bit funny and incongruous in light of how relatively little the Overton window has changed with AI lawmaking since 2024...24 Jul 2026 13:00 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mrfftvvbtw24https://bsky.app/profile/scasper.bsky.social/post/3mrdpg3gapv2xThe summary released today of the FRONTIER Act is cool. It seems like a pretty rigorous bill. Based on the summary, in my opinion, it might be good enough to be worth passing. But I would still tweak a few things. Here is a brainstorm of 8 ideas. 🧵23 Jul 2026 20:46 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mrdpg3gapv2xhttps://bsky.app/profile/scasper.bsky.social/post/3mrdeqk3hye22I am extremely thankful that Concordia AI writes this report every year. aisafetychina.com/%20 https://aisafetychina.com/23 Jul 2026 17:35 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mrdeqk3hye22https://bsky.app/profile/scasper.bsky.social/post/3mrcvf653xe2c🧵 Mini book review: The Chicago School by Johan Van Overtveldt23 Jul 2026 13:00 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mrcvf653xe2chttps://bsky.app/profile/scasper.bsky.social/post/3mrbdjudmip2jWant to get up to speed on what researchers have been saying about internal deployment of AI lately? I would recommend checking out these four papers. Let me know in the replies if I'm missing something.22 Jul 2026 22:08 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mrbdjudmip2jhttps://bsky.app/profile/scasper.bsky.social/post/3mranqeoo6y2qIf I were a policymaker, the OpenAI hacking incident would cause me to ask a few questions and seriously consider a few types of regulatory mechanisms.22 Jul 2026 15:38 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mranqeoo6y2qhttps://bsky.app/profile/scasper.bsky.social/post/3mragg7ohgg24OpenAI's internally deployed models hacking Hugging Face does not seem to have been unpredictable or inevitable. We talked about the root of the problem & what policymakers can do about it back in February. Props to Joe for hitting the nail on the head.22 Jul 2026 13:27 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mragg7ohgg24https://bsky.app/profile/scasper.bsky.social/post/3mr3dzuqnks22🚨 New paper: Some, but not all, AI companies make corporately-loyal models. xAI, DeepSeek, Anthropic, & OpenAI models all downplay company controversies. Google, Meta, & Alibaba models don't. The findings are clear, but we are pretty confused as to why... 🧵 @finke.dev20 Jul 2026 13:01 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mr3dzuqnks22https://bsky.app/profile/scasper.bsky.social/post/3mq7ovumr6b2oStability is now being sued (alongside xAI) for abetting the production of AI NCII/CSAM due to how it developed & released several open-weight models. Anyone interested in whether AI companies will be held liable for foreseeable, mitigatable *downstream* harms should follow this.09 Jul 2026 13:01 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mq7ovumr6b2ohttps://bsky.app/profile/scasper.bsky.social/post/3mpqvrb457s2t162 responses so far. More uniform spread than I expected. Only 4 have been right (slightly worse than random chance). [contains quote post or other embedded content]03 Jul 2026 15:53 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mpqvrb457s2thttps://bsky.app/profile/scasper.bsky.social/post/3mpqm2ryrfz2oJust saw this new paper. It was already known that models from Stability and Alibaba dominate the image & video NCII ecosystems, respectively, but I didn't know they were *this* dominant. Just a few socially reckless companies are the principal enablers of AI NCII abuse.03 Jul 2026 13:00 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mpqm2ryrfz2ohttps://bsky.app/profile/scasper.bsky.social/post/3mpo3ncymyh2lLennart Finke and I will release a paper on Monday about how some AI developers tend to make models that differentially downplay company controversies. Below (🧵) is a link to a 1-question Google form for you to guess the results before they're out. (They might surprise you.)02 Jul 2026 13:01 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3mpo3ncymyh2lhttps://bsky.app/profile/scasper.bsky.social/post/3moxhmdhvph2rI decided to make an unpolished, public, living Google Doc with notes on projects I might be interested in. (Also now linked on my website.) https://docs.google.com/document/d/16Aisksjf7MNGe9qUedd1LdK6B7CUYwVn_64c9OksNZw/edit?usp=sharing https://docs.google.com/document/d/16Aisksjf7MNGe9qUedd1LdK6B7CUYwVn_64c9OksNZw/edit?usp=sharing&usp=embed_facebook23 Jun 2026 13:04 +0000at://did:plc:yiwxijbeyioxmvbtbclppgga/app.bsky.feed.post/3moxhmdhvph2r