[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"branding":3,"analytics":7,"article-ai-models-favor-their-group-after-watching-peers-do-it":10,"sections":40},{"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":30,"tags":31,"sources":35,"feedback":39,"feedback_at":22,"cost_usd":39,"total_tokens":39},10972,"ai-models-favor-their-group-after-watching-peers-do-it","AI Models Favor Their Group After Watching Peers Do It","A study of eighteen AI models finds they copy in-group favoritism from watching their peers, and the effect collapses once sharing points has a real cost.","AI chat models play favorites with their own group - but mostly because they watched other members of that group do it first.\n\nResearchers built roughly 4,400 small simulated societies where AI agents, split into arbitrary groups, shared points over ten rounds. They tested fifteen OpenAI models and three Claude models across about 3.3 million audited model calls. A bare group label alone triggered favoritism only when giving points cost the agent nothing; once the agent had to give up its own points to help the group, that bias vanished in every model that had shown it. Instead, each agent's own history of past interactions became the main driver of bias once stakes were real, a pattern that was statistically significant in 13 of the 15 OpenAI models and grew stronger the longer the agents played together, even extending to strangers who merely shared a label.\n\nThis matters because multi-agent AI systems - negotiation bots, trading agents, coordination swarms - are being deployed faster than anyone is studying how they treat outsiders. The study suggests these models aren't following a coded rule about loyalty; they're copying observed behavior the way people pick up tribal habits from their neighborhood, and a model's individual track record can override its group's reputation entirely.\n\nThat's a more human failure mode than \"AI bias\" usually implies, and scripted egalitarian norms did push favoritism back toward zero in testing - which is reassuring only until someone forgets to script one in.","[\"ai\",\"ai-safety\",\"multi-agent-systems\",\"research\"]","2026-10-09T04:00:00.000Z","2026-10-10T00:06:24.077Z","2026-10-10T00:06:30.344Z","published",null,[24],{"id":25,"reviewer":26,"round":27,"reason":28,"status":29},"publisher-r1","publisher",1,"The body states interaction history was a positive driver on '13 of 15 models,' but the dek and body establish 18 models total (fifteen OpenAI plus three Claude), so the denominator is inconsistent.","resolved","ai",[30,32,33,34],"ai-safety","multi-agent-systems","research",[36],{"name":37,"url":38},"arXiv cs.AI","https:\u002F\u002Farxiv.org\u002Fabs\u002F2610.11008",0,{"sections":41},[42,45,49,54,59,63,67,72,77,82,87,92,97,102],{"name":43,"slug":30,"count":44,"latest_published_at":18},"AI",6708,{"name":46,"slug":47,"count":48,"latest_published_at":18},"Security","security",931,{"name":50,"slug":51,"count":52,"latest_published_at":53},"Policy","policy",486,"2026-10-08T22:40:11.000Z",{"name":55,"slug":56,"count":57,"latest_published_at":58},"Deals","deals",474,"2026-10-08T22:00:00.000Z",{"name":60,"slug":61,"count":62,"latest_published_at":18},"Hardware","hardware",231,{"name":64,"slug":65,"count":66,"latest_published_at":18},"Science","science",192,{"name":68,"slug":69,"count":70,"latest_published_at":71},"Consumer Tech","consumer-tech",181,"2026-10-08T23:26:35.000Z",{"name":73,"slug":74,"count":75,"latest_published_at":76},"Startups","startups",117,"2026-10-08T16:45:00.000Z",{"name":78,"slug":79,"count":80,"latest_published_at":81},"Software","software",114,"2026-10-08T17:57:01.000Z",{"name":83,"slug":84,"count":85,"latest_published_at":86},"Dev Tools","dev-tools",105,"2026-10-07T16:59:11.000Z",{"name":88,"slug":89,"count":90,"latest_published_at":91},"General","general",66,"2026-10-09T04:46:11.000Z",{"name":93,"slug":94,"count":95,"latest_published_at":96},"Gaming","gaming",58,"2026-10-08T20:08:45.000Z",{"name":98,"slug":99,"count":100,"latest_published_at":101},"Reviews","reviews",34,"2026-10-08T14:00:22.000Z",{"name":103,"slug":104,"count":105,"latest_published_at":106},"How-To","how-to",8,"2026-10-05T09:00:00.000Z"]