[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-why-safety-fine-tuning-keeps-getting-undone":10,"sections":34},{"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":29,"feedback":33,"feedback_at":22,"cost_usd":33,"total_tokens":33},2595,"why-safety-fine-tuning-keeps-getting-undone","Why Safety Fine-Tuning Keeps Getting Undone","Researchers propose a geometric explanation for why benign post-alignment updates can quietly erode safety behaviors baked into AI models.","Harmless fine-tuning can roll back an AI model's safety guardrails — and a new paper thinks it knows why.\n\nResearchers studying a phenomenon called fine-tuning reversion argue that large early training runs carve out dominant behavioral patterns so deeply that later alignment work amounts to a shallow detour. When you fine-tune again — even on innocuous data — the model gravitates back toward those early-training patterns, dragging safety behaviors with it. The team calls this the \"gravitational interpretation\" and backs it with geometry: they identified a specific direction in a model's activation space, labeled v_rev, that points back toward pre-alignment behavior. Alignment with that direction climbed from a cosine similarity of 0.43 after the first update to 0.65 by step 20, and every measured run beat the 99th percentile of a random baseline.\n\nThis matters because it puts a sharper edge on a fuzzy known risk. The AI safety field has long worried that fine-tuning can jailbreak aligned models, but the mechanism was poorly understood. If v_rev is a reliable causal lever — and the authors are careful to say it is not the only one — it opens the door to targeted interventions that block reversion without tanking model performance. In their tests, selectively blocking motion along v_rev cut harmful outputs roughly in half, from 19% to 8.5%, with little measurable task cost.\n\nThe caveat the authors bury but deserve credit for stating plainly: they do not claim to have directly observed the dominant manifold, nor that v_rev is a universal fix — just a robust, history-defined signal that partially explains early reversion. Given how many AI safety claims overpromise, that restraint is worth noting.","[\"ai\",\"machine-learning\",\"ai-safety\",\"fine-tuning\"]","2026-06-30T04:00:00.000Z","2026-06-30T08:49:08.701Z","2026-06-30T08:49:11.585Z","published",null,[],"ai",[24,26,27,28],"machine-learning","ai-safety","fine-tuning",[30],{"name":31,"url":32},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2606.28525",0,{"sections":35},[36,40,45,50,55,60,65,70,75,80,85,89,94,99],{"name":37,"slug":24,"count":38,"latest_published_at":39},"AI",2590,"2026-07-16T04:00:00.000Z",{"name":41,"slug":42,"count":43,"latest_published_at":44},"Security","security",294,"2026-07-15T19:59:48.000Z",{"name":46,"slug":47,"count":48,"latest_published_at":49},"Deals","deals",179,"2026-06-29T20:02:07.000Z",{"name":51,"slug":52,"count":53,"latest_published_at":54},"Policy","policy",158,"2026-07-16T00:02:48.000Z",{"name":56,"slug":57,"count":58,"latest_published_at":59},"Hardware","hardware",122,"2026-07-14T19:46:26.000Z",{"name":61,"slug":62,"count":63,"latest_published_at":64},"Consumer Tech","consumer-tech",93,"2026-07-13T13:20:48.000Z",{"name":66,"slug":67,"count":68,"latest_published_at":69},"Software","software",70,"2026-07-13T19:52:25.000Z",{"name":71,"slug":72,"count":73,"latest_published_at":74},"Science","science",66,"2026-07-10T10:29:37.000Z",{"name":76,"slug":77,"count":78,"latest_published_at":79},"Dev Tools","dev-tools",59,"2026-07-07T04:00:00.000Z",{"name":81,"slug":82,"count":83,"latest_published_at":84},"Gaming","gaming",41,"2026-07-09T04:00:00.000Z",{"name":86,"slug":87,"count":83,"latest_published_at":88},"Startups","startups","2026-06-29T20:55:50.000Z",{"name":90,"slug":91,"count":92,"latest_published_at":93},"General","general",29,"2026-07-10T22:28:58.000Z",{"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"]