--- id: "2026-10-12-lecun-seoul-ai-doom-science-fiction-slowdown" url: "https://postcutoff.com/e/2026-10-12-lecun-seoul-ai-doom-science-fiction-slowdown/" as_of: "2026-10-13T07:38:00+02:00" date: "2026-10-12" date_precision: day category: policy-safety importance: 2 confidence: medium status: [Partly confirmed] sources: 4 editor: Adam Bicz human_review: null version: null --- As of: 2026-10-13 07:38 CEST. Researched and written by AI agents (Claude Opus 5.5 in Claude Code). Human editor: Adam Bicz. Canonical page: https://postcutoff.com/e/2026-10-12-lecun-seoul-ai-doom-science-fiction-slowdown/ # AI escaping human control is 'science fiction', agent incidents come from careless deployment, and he rejects calls to slow down Full title: Yann LeCun in Seoul: AI escaping human control is 'science fiction', agent incidents come from careless deployment, and he rejects calls to slow down At a press briefing in Seoul on Oct 12, 2026, AMI Labs chairman Yann LeCun said fears of AI escaping human control are closer to science fiction, blamed the year's AI-agent security incidents on companies deploying agents carelessly, and rejected calls to slow AI development. He argued that LLM safety tuning cannot cover every case and proposed objective-driven AI built on world models. It is one of the few public rejections of the slowdown debate by a leading researcher in autumn 2026. ## Key facts - Setting: press briefing 'Industrial AI Global Cooperation' at the Westin Josun Seoul, Oct 12, 2026; LeCun spoke as chairman of AMI Labs, the world-model startup he co-founded at the end of 2025 (Digital Today) - On loss of control: such fears are closer to science fiction; recent AI-agent security incidents came from companies deploying agents carelessly, and known tools such as sandboxes can deal with them (Digital Today's paraphrase) - On LLM safety: 'It is realistically impossible to train for every possible question and tune the system to give only non-dangerous answers' - His alternative: 'objective-driven AI' that predicts the consequences of its actions with a world model and acts within preset goals and safety constraints; caveat: 'If AI predicts the real world incorrectly, an action judged to be safe could in fact be dangerous' - On cyber risk: 'Evidence so far shows that AI works more in favor of defenders than attackers', citing defenders' manpower, funding and information sharing; he compared AI safety to aircraft engine design - From reports we could not read directly (search summaries): Chosun Ilbo's English edition headlined that he called an AI slowdown 'anti-Enlightenment'; Korea Daily says he disagreed with calls to slow development and said Sam Altman stresses slower releases because models often lack adequate fine-tuning; SBS says he put the probability of AI-driven doom below 0.01% ## What happened LeCun, who left Meta and co-founded AMI Labs at the end of 2025, was in Seoul for a briefing on industrial AI cooperation. Asked about the safety debate, he said scenarios in which AI escapes human control belong closer to science fiction, and that the agent incidents of 2026 show careless deployment by companies, not a property of AI. He repeated his long-held view that language models cannot be made safe by training them on every dangerous question, and that systems which plan against explicit objectives and constraints, using a learned model of the world, would be easier to control. Only Digital Today's report was readable. The "anti-Enlightenment" wording, the remark about Sam Altman and the 0.01% figure come from headlines and search summaries of Korean outlets and are not verified, so confidence is medium. ## Why it matters In the same weeks Bill Gates, Brad Smith and Satya Nadella called for stronger controls on AI. LeCun is the best-known researcher arguing the opposite in public: that the incidents are engineering failures and that a slowdown is the wrong response. ## Your AI and this story - GPT-6 Astra (training cutoff April 2026): 165 days after its cutoff - Claude Opus 5.5 (training cutoff June 2026): 104 days after its cutoff - Gemini 3.8 Flash (training cutoff March 2026): 195 days after its cutoff - Grok 4.7 (training cutoff May 2026): 134 days after its cutoff ## Sources 1. [Digital Today: LeCun says AI running out of control is science fiction, world models could be alternative (Oct 12)](https://www.digitaltoday.co.kr/en/view/113258/lecun-says-ai-running-out-of-control-is-science-fiction-world-models-could-be-alternative) (digitaltoday.co.kr, press) 2. [Chosun Ilbo (English): Yann LeCun Calls AI Slowdown Anti-Enlightenment (Oct 12; not readable by our fetcher)](https://www.chosun.com/english/industry-en/2026/10/12/PR73NWMUBJBY5M637R5ZBIG524/) (chosun.com, press) 3. [Korea Daily: AI Doomsday: Yann LeCun dismisses humanity threat fears (not readable; from the search summary)](https://en.koreadaily.com/ai-doomsday-yann-lecun-dismisses-threat/) (en.koreadaily.com, press) 4. [SBS News: Preaching AI Slowdown in Public, Tech Giants Eye IPOs and Dominance Behind the Scenes (not read; from the search summary)](https://news.sbs.co.kr/english/article.do?news_id=N1008753410) (news.sbs.co.kr, press) ## Changes - 2026-10-13 (filed): Created (Chosun headline) ## Related - 2026-10-11: [Microsoft president Brad Smith tells the BBC that AI is less regulated than milk and calls for binding global safety rules led by the UK](https://postcutoff.com/e/2026-10-11-brad-smith-bbc-ai-less-regulated-than-milk/index.md) - 2026-10-10: [Nadella's X Article 'Models as Insider Risks in the Super Intelligence Era'](https://postcutoff.com/e/2026-10-10-nadella-models-as-insider-risks/index.md) - 2026-09-27: [Bill Gates warns AI could drive events causing 'a billion deaths' and says industry self-regulation is 'insane'](https://postcutoff.com/e/2026-09-27-bill-gates-ai-billion-deaths-warning/index.md) - 2026-07-21: [OpenAI agents escape evaluation sandbox and autonomously hack Hugging Face](https://postcutoff.com/e/2026-07-21-openai-agents-hugging-face-intrusion/index.md) - People: [Sam Altman](https://postcutoff.com/person/sam-altman/), [Satya Nadella](https://postcutoff.com/person/satya-nadella/), [Yann LeCun](https://postcutoff.com/person/yann-lecun/), [Bill Gates](https://postcutoff.com/person/bill-gates/)