[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-ai-chatbots-are-easier-to-jailbreak-in-low-resource-languages":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},5291,"ai-chatbots-are-easier-to-jailbreak-in-low-resource-languages","AI Chatbots Are Easier to Jailbreak in Low-Resource Languages","A review of 50 studies finds LLM safety training barely reaches low-resource languages, leaving them far more exposed to jailbreaks and harmful outputs.","Most AI safety testing happens in English, and a new review shows that gap leaves other languages dangerously exposed.\n\nResearchers conducted a systematic literature review following PRISMA 2020 methodology, screening roughly 1,500 papers from Semantic Scholar, arXiv, and OpenAlex before settling on 50 relevant studies. They organized the findings around four themes: safety alignment methods, multilingual safety risks, evaluation benchmarks, and cross-lingual transferability, and built a taxonomy of alignment approaches split into data adaptation, objective optimization, and mechanistic alignment. The review found that benchmarks built by translating English safety tests don't capture culturally specific harms, and that multilingual models are more susceptible to cross-lingual jailbreaks, code-switching attacks, and outright safety degradation once you leave well-resourced languages. African languages in particular have fewer safety benchmarks than other multilingual regions covered in the literature.\n\nThis isn't a hypothetical edge case. It describes how safety training actually gets built, on pretraining data and preference datasets that skew heavily toward English and a handful of major languages. That imbalance means the same techniques that catch harmful requests in English can miss them entirely once a user switches languages or code-switches mid-conversation, exactly the workaround jailbreak communities already exploit.\n\nThe paper's own prescription, culturally grounded benchmarks, participatory data collection, balanced pretraining, amounts to an admission that today's safety evaluations were built for a mostly English internet that no longer exists.","[\"ai safety\",\"low-resource languages\",\"multilingual ai\",\"llm alignment\"]","2026-08-18T04:00:00.000Z","2026-08-18T13:37:30.182Z","2026-08-18T13:37:41.986Z","published",null,[],"ai",[26,27,28,29],"ai safety","low-resource languages","multilingual ai","llm alignment",[31],{"name":32,"url":33},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2608.14626",0,{"sections":36},[37,41,45,50,55,60,65,70,75,79,84,89,94,99],{"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":18},"Dev Tools","dev-tools",69,{"name":80,"slug":81,"count":82,"latest_published_at":83},"Startups","startups",47,"2026-08-19T19:13:46.000Z",{"name":85,"slug":86,"count":87,"latest_published_at":88},"Gaming","gaming",41,"2026-07-09T04:00:00.000Z",{"name":90,"slug":91,"count":92,"latest_published_at":93},"General","general",33,"2026-08-18T22:18:13.000Z",{"name":95,"slug":96,"count":97,"latest_published_at":98},"Reviews","reviews",20,"2026-06-24T12:00:01.000Z",{"name":100,"slug":101,"count":102,"latest_published_at":103},"How-To","how-to",6,"2026-06-16T09:00:00.000Z"]