[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-multilingual-ai-safety-gap-traced-to-a-late-refusal-circuit":10,"sections":35},{"siteName":4,"siteTagline":5,"publisherName":4,"contactEmail":6},"The Revision","Tech news, decoded.","editor@therevision.news",{"gaMeasurementId":8,"adsenseClientId":9},"G-ZW2MV82GYR","ca-pub-8533917693782264",{"article":11},{"id":12,"slug":13,"title":14,"dek":15,"body_md":16,"tags_json":17,"published_at":18,"created_at":19,"updated_at":20,"status":21,"review_note":22,"review_notes":23,"image_url":22,"persona_id":22,"persona_name":22,"section":24,"tags":25,"sources":30,"feedback":34,"feedback_at":22,"cost_usd":34,"total_tokens":34},5704,"multilingual-ai-safety-gap-traced-to-a-late-refusal-circuit","Multilingual AI Safety Gap Traced to a Late Refusal Circuit","A new study finds multilingual AI models detect harmful requests early but write refusals late, making the safety gap costly to fix in low-resource languages.","A safety flaw in multilingual AI models turns out to be a timing problem, not a comprehension one.\n\nResearchers studied sarvam, an Indic-multilingual mixture-of-experts (MoE) model that only activates the sub-networks it needs for a given prompt. They found the model recognizes a harmful request just as well in Hindi or other Indic languages as it does in English; the internal signal for 'this is harmful' barely changes across languages. The gap shows up later: refusing to answer isn't decided in that single moment of recognition, but assembled gradually as the model generates its response. The researchers traced that refusal-writing process to a specific circuit, where one component pushes toward refusal and another, an 'opposer,' holds it back, and in lower-resource languages, the opposer wins more often.\n\nThat distinction matters for anyone trying to patch these safety gaps. Turning down the opposing signal fixed the cross-language refusal gap cheaply; forcing the refusal signal itself to fire harder was far more expensive computationally; and editing the specific attention components by hand did essentially nothing. It's a rare case in AI safety research where a fix comes with an actual price tag attached, not just a diagnosis.\n\nThe same circuit pattern turned up in a second, unrelated MoE model, so this isn't just a sarvam quirk. But the researchers note the cost of each fix is architecture-specific, which means there's no single cheap patch every multilingual model can borrow.","[\"ai-safety\",\"multilingual-models\",\"mixture-of-experts\",\"llm-alignment\"]","2026-08-19T04:00:00.000Z","2026-08-19T13:07:41.782Z","2026-08-19T13:07:53.686Z","published",null,[],"ai",[26,27,28,29],"ai-safety","multilingual-models","mixture-of-experts","llm-alignment",[31],{"name":32,"url":33},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2608.08032",0,{"sections":36},[37,41,45,50,55,60,65,70,75,80,85,90,95,100],{"name":38,"slug":24,"count":39,"latest_published_at":40},"AI",3293,"2026-08-20T04:00:00.000Z",{"name":42,"slug":43,"count":44,"latest_published_at":40},"Security","security",435,{"name":46,"slug":47,"count":48,"latest_published_at":49},"Policy","policy",210,"2026-08-19T09:32:27.000Z",{"name":51,"slug":52,"count":53,"latest_published_at":54},"Deals","deals",179,"2026-06-29T20:02:07.000Z",{"name":56,"slug":57,"count":58,"latest_published_at":59},"Hardware","hardware",140,"2026-08-19T18:25:42.000Z",{"name":61,"slug":62,"count":63,"latest_published_at":64},"Consumer Tech","consumer-tech",95,"2026-08-18T16:05:00.000Z",{"name":66,"slug":67,"count":68,"latest_published_at":69},"Science","science",90,"2026-08-19T18:41:02.000Z",{"name":71,"slug":72,"count":73,"latest_published_at":74},"Software","software",73,"2026-08-18T07:51:50.000Z",{"name":76,"slug":77,"count":78,"latest_published_at":79},"Dev Tools","dev-tools",69,"2026-08-18T04:00:00.000Z",{"name":81,"slug":82,"count":83,"latest_published_at":84},"Startups","startups",47,"2026-08-19T19:13:46.000Z",{"name":86,"slug":87,"count":88,"latest_published_at":89},"Gaming","gaming",41,"2026-07-09T04:00:00.000Z",{"name":91,"slug":92,"count":93,"latest_published_at":94},"General","general",33,"2026-08-18T22:18:13.000Z",{"name":96,"slug":97,"count":98,"latest_published_at":99},"Reviews","reviews",20,"2026-06-24T12:00:01.000Z",{"name":101,"slug":102,"count":103,"latest_published_at":104},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]