[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-researchers-booby-trap-llm-weights-to-block-jailbreak-edits":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},6219,"researchers-booby-trap-llm-weights-to-block-jailbreak-edits","Researchers Booby-Trap LLM Weights to Block Jailbreak Edits","A new technique poisons the internal signals attackers use to edit open-weight model weights, cutting successful jailbreak edits dramatically in tests.","A new paper shows how to fight AI jailbreaks by lying to the attacker's own measurement tools.\n\nResearchers describe representation engineering attacks that let anyone with a single GPU estimate an open-weight model's \"refusal direction\" and edit its weights to suppress safety training in minutes, with no gradient-based training required. Their defense, called Bait-and-Recover, plants a \"bait\" adapter at the layer attackers read and a paired \"recovery\" adapter one layer later, trained with gradient routing to split what the model shows from what it actually does. The bait layer feeds attackers a poisoned signal that throws off their edit search, while the recovery layer quietly restores normal computation downstream. Tested across four open-weight models under a strict limit on how much the edits could change outputs, the defense raised the minimum refusal rate against these attacks from 16.25% to 71.75%, with negligible impact on general benchmarks.\n\nMost LLM safety work assumes attackers only get black-box access, a chat window, not the weights. Open-weight releases blow up that assumption, and this paper's baseline numbers show how flimsy current alignment is once someone can inspect activations directly: a 16% refusal floor means these edits work on most attempts. Bait-and-Recover does not close that hole so much as make the attacker's own toolkit lie to them, a cheaper fix than retraining a model's values from scratch.\n\nIt is also a reminder that \"safety-tuned\" and \"safe to publish the weights of\" are not the same claim, whatever a model card says.","[\"ai safety\",\"jailbreak defense\",\"open-weight models\",\"llm security\"]","2026-09-10T04:00:00.000Z","2026-09-10T06:44:41.468Z","2026-09-10T06:44:53.344Z","published",null,[],"security",[26,27,28,29],"ai safety","jailbreak defense","open-weight models","llm security",[31],{"name":32,"url":33},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2609.05794",0,{"sections":36},[37,42,45,50,55,60,65,69,74,79,84,89,94,99],{"name":38,"slug":39,"count":40,"latest_published_at":41},"AI","ai",3480,"2026-09-11T04:00:00.000Z",{"name":43,"slug":24,"count":44,"latest_published_at":41},"Security",629,{"name":46,"slug":47,"count":48,"latest_published_at":49},"Policy","policy",336,"2026-09-11T00:56:21.000Z",{"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",153,"2026-09-09T15:12:32.000Z",{"name":61,"slug":62,"count":63,"latest_published_at":64},"Consumer Tech","consumer-tech",99,"2026-09-09T17:27:33.000Z",{"name":66,"slug":67,"count":68,"latest_published_at":41},"Science","science",98,{"name":70,"slug":71,"count":72,"latest_published_at":73},"Software","software",75,"2026-09-10T20:41:21.000Z",{"name":75,"slug":76,"count":77,"latest_published_at":78},"Dev Tools","dev-tools",69,"2026-08-18T04:00:00.000Z",{"name":80,"slug":81,"count":82,"latest_published_at":83},"Startups","startups",55,"2026-09-09T23:14:29.000Z",{"name":85,"slug":86,"count":87,"latest_published_at":88},"Gaming","gaming",43,"2026-09-10T12:18:06.000Z",{"name":90,"slug":91,"count":92,"latest_published_at":93},"General","general",41,"2026-09-08T01:57:23.000Z",{"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"]