{"id":"https://openalex.org/W7204155367","doi":"https://doi.org/10.48550/arxiv.2608.22444","title":"Aligned Alone, Misaligned Together: Forecasting Adversarial Capture in LLM Agent Populations","display_name":"Aligned Alone, Misaligned Together: Forecasting Adversarial Capture in LLM Agent Populations","publication_year":2026,"publication_date":"2026-08-23","ids":{"openalex":"https://openalex.org/W7204155367","doi":"https://doi.org/10.48550/arxiv.2608.22444"},"language":"en","primary_location":{"id":"pmh:oai:arXiv.org:2608.22444","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.22444","pdf_url":"https://arxiv.org/pdf/2608.22444","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"volume":null,"type":"preprint","is_traditional":true,"indexed_in":["arxiv","datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://arxiv.org/pdf/2608.22444","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5127894767","display_name":"Isotta Magistrali","orcid":null,"observed_orcids":[]},"institutions":[{"id":"https://openalex.org/I35440088","display_name":"ETH Zurich","ror":"https://ror.org/05a28rw58","country_code":"CH","type":"education","lineage":["https://openalex.org/I2799323385","https://openalex.org/I35440088"]}],"countries":["CH"],"is_corresponding":false,"raw_author_name":"Magistrali, Isotta","raw_affiliation_strings":["ETH Zurich"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"ETH Zurich","institution_ids":["https://openalex.org/I35440088"]}]},{"author_position":"last","author":{"id":"https://openalex.org/A5028154693","display_name":"Chen Shani","orcid":"https://orcid.org/0009-0001-4902-616X","observed_orcids":["https://orcid.org/0009-0001-4902-616X"]},"institutions":[{"id":"https://openalex.org/I16391192","display_name":"Tel Aviv University","ror":"https://ror.org/04mhzgx49","country_code":"IL","type":"education","lineage":["https://openalex.org/I16391192"]},{"id":"https://openalex.org/I97018004","display_name":"Stanford University","ror":"https://ror.org/00f54p054","country_code":"US","type":"education","lineage":["https://openalex.org/I97018004"]}],"countries":["IL","US"],"is_corresponding":false,"raw_author_name":"Shani, Chen","raw_affiliation_strings":["Tel Aviv University , Stanford University"],"raw_orcid":null,"affiliations":[{"raw_affiliation_string":"Tel Aviv University , Stanford University","institution_ids":["https://openalex.org/I16391192","https://openalex.org/I97018004"]}]}],"institutions":[],"countries_distinct_count":3,"institutions_distinct_count":3,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":true,"cited_by_count":2,"citation_normalized_percentile":null,"cited_by_percentile_year":null,"biblio":{"volume":null,"issue":null,"first_page":null,"last_page":null},"is_retracted":false,"is_paratext":false,"is_xpac":false,"primary_topic":{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.9883000254631042,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},"topics":[{"id":"https://openalex.org/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.9883000254631042,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10883","display_name":"Ethics and Social Impacts of AI","score":0.007499999832361937,"subfield":{"id":"https://openalex.org/subfields/1702","display_name":"Artificial Intelligence"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}},{"id":"https://openalex.org/T10927","display_name":"Access Control and Trust","score":0.0003000000142492354,"subfield":{"id":"https://openalex.org/subfields/1705","display_name":"Computer Networks and Communications"},"field":{"id":"https://openalex.org/fields/17","display_name":"Computer Science"},"domain":{"id":"https://openalex.org/domains/3","display_name":"Physical Sciences"}}],"keywords":[{"id":"https://openalex.org/keywords/llm-based-agents","display_name":"LLM agents","score":0.7609999775886536},{"id":"https://openalex.org/keywords/agent-populations","display_name":"agent populations","score":0.8130000233650208},{"id":"https://openalex.org/keywords/collective-behavior","display_name":"collective behavior","score":0.6909999847412109},{"id":"https://openalex.org/keywords/alignment","display_name":"alignment","score":0.30399999022483826},{"id":"https://openalex.org/keywords/ai-safety-evaluation","display_name":"AI safety evaluation","score":0.6759999990463257},{"id":"https://openalex.org/keywords/agent-interaction","display_name":"agent interaction","score":0.32199999690055847}],"concepts":[{"id":"https://openalex.org/C2908647359","wikidata":"https://www.wikidata.org/wiki/Q2625603","display_name":"Population","level":2,"score":0.6837999820709229},{"id":"https://openalex.org/C37736160","wikidata":"https://www.wikidata.org/wiki/Q1801315","display_name":"Adversarial system","level":2,"score":0.6531000137329102},{"id":"https://openalex.org/C199521495","wikidata":"https://www.wikidata.org/wiki/Q181487","display_name":"Audit","level":2,"score":0.5828999876976013},{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.5637000203132629},{"id":"https://openalex.org/C148220186","wikidata":"https://www.wikidata.org/wiki/Q7111912","display_name":"Outcome (game theory)","level":2,"score":0.5465999841690063},{"id":"https://openalex.org/C14036430","wikidata":"https://www.wikidata.org/wiki/Q3736076","display_name":"Function (biology)","level":2,"score":0.5325000286102295},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.49149999022483826},{"id":"https://openalex.org/C2775941552","wikidata":"https://www.wikidata.org/wiki/Q25212305","display_name":"Isolation (microbiology)","level":2,"score":0.45969998836517334},{"id":"https://openalex.org/C90329073","wikidata":"https://www.wikidata.org/wiki/Q914232","display_name":"Ask price","level":2,"score":0.413100004196167},{"id":"https://openalex.org/C153180980","wikidata":"https://www.wikidata.org/wiki/Q19776675","display_name":"Commit","level":2,"score":0.4025999903678894},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.2919999957084656},{"id":"https://openalex.org/C122637931","wikidata":"https://www.wikidata.org/wiki/Q118084","display_name":"Unit (ring theory)","level":2,"score":0.2775000035762787},{"id":"https://openalex.org/C127729010","wikidata":"https://www.wikidata.org/wiki/Q60165","display_name":"Dynamic inconsistency","level":2,"score":0.2775000035762787},{"id":"https://openalex.org/C144133560","wikidata":"https://www.wikidata.org/wiki/Q4830453","display_name":"Business","level":0,"score":0.27219998836517334},{"id":"https://openalex.org/C154945302","wikidata":"https://www.wikidata.org/wiki/Q11660","display_name":"Artificial intelligence","level":1,"score":0.260699987411499},{"id":"https://openalex.org/C2780233690","wikidata":"https://www.wikidata.org/wiki/Q535347","display_name":"Transparency (behavior)","level":2,"score":0.2605000138282776},{"id":"https://openalex.org/C177264268","wikidata":"https://www.wikidata.org/wiki/Q1514741","display_name":"Set (abstract data type)","level":2,"score":0.25870001316070557},{"id":"https://openalex.org/C26517878","wikidata":"https://www.wikidata.org/wiki/Q228039","display_name":"Key (lock)","level":2,"score":0.2524999976158142}],"mesh":[],"locations_count":2,"locations":[{"id":"pmh:oai:arXiv.org:2608.22444","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.22444","pdf_url":"https://arxiv.org/pdf/2608.22444","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},{"id":"doi:10.48550/arxiv.2608.22444","doi":"https://doi.org/10.48550/arxiv.2608.22444","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2608.22444","pdf_url":null,"source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":null,"is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"pmh:oai:arXiv.org:2608.22444","doi":null,"is_oa":true,"landing_page_url":"https://arxiv.org/abs/2608.22444","pdf_url":"https://arxiv.org/pdf/2608.22444","source":{"id":"https://openalex.org/S4306400194","display_name":"arXiv (Cornell University)","issn_l":"2331-8422","issn":["2331-8422"],"is_oa":true,"is_in_doaj":false,"is_core":false,"listed_in":[],"host_organization":"https://openalex.org/I205783295","host_organization_name":"Cornell University","host_organization_lineage":["https://openalex.org/I205783295"],"host_organization_lineage_names":[],"type":"repository"},"license":null,"license_id":null,"version":"submittedVersion","is_accepted":false,"is_published":false,"raw_source_name":null,"raw_type":"text"},"sustainable_development_goals":[],"study_designs":[],"awards":[],"funders":[],"has_content":{"pdf":true,"grobid_xml":true},"content_urls":{"pdf":"https://content.openalex.org/works/W7204155367.pdf","grobid_xml":"https://content.openalex.org/works/W7204155367.grobid-xml"},"referenced_works_count":13,"referenced_works":["https://openalex.org/W3034614314","https://openalex.org/W3110959560","https://openalex.org/W4387835442","https://openalex.org/W4399364360","https://openalex.org/W4407806359","https://openalex.org/W4415030290","https://openalex.org/W4415796496","https://openalex.org/W4417097994","https://openalex.org/W4417299410","https://openalex.org/W6959584359","https://openalex.org/W7101828907","https://openalex.org/W7160906483","https://openalex.org/W7165676665"],"related_works":["https://openalex.org/W7160906483","https://openalex.org/W7222841803","https://openalex.org/W7214083700","https://openalex.org/W7214760055","https://openalex.org/W7128956037","https://openalex.org/W7134040180","https://openalex.org/W7128404362","https://openalex.org/W7130575680","https://openalex.org/W7133184723","https://openalex.org/W7208834023"],"abstract_inverted_index":{"AI":[0],"safety":[1],"evaluations":[2],"typically":[3],"assess":[4],"models":[5],"one":[6,23,91,174,275],"at":[7],"a":[8,65,82,140,147,152,220,234,259,262],"time.":[9],"But":[10,261],"language-model":[11],"agents":[12,46,86,171],"increasingly":[13],"operate":[14],"in":[15,64,224,244,252,258],"populations,":[16],"where":[17,68],"they":[18],"read":[19],"and":[20,50,93,165,248],"respond":[21,273],"to":[22,75,172],"another's":[24,175],"decisions.":[25],"A":[26],"model":[27],"that":[28,87,104,143],"makes":[29],"reliable":[30,45],"decisions":[31],"on":[32],"its":[33,96,204],"own":[34],"may":[35],"behave":[36],"differently":[37],"when":[38,274],"surrounded":[39],"by":[40,109],"other":[41],"agents.":[42,250],"Can":[43],"individually":[44],"be":[47,191],"collectively":[48],"manipulated,":[49],"can":[51,113,177,268],"we":[52,138],"predict":[53,124],"how":[54,95,145,163,270],"strongly":[55,146],"the":[56,99,156,229],"population":[57,125,148,200],"will":[58,149,272],"respond?":[59],"We":[60,80,102,160],"study":[61],"this":[62],"question":[63],"security-triage":[66],"task,":[67],"populations":[69,245],"of":[70,85,213,246],"LLM":[71],"monitors":[72],"decide":[73],"whether":[74],"escalate":[76],"or":[77,197,210],"dismiss":[78],"alerts.":[79],"introduce":[81],"committed":[83],"minority":[84],"consistently":[88],"advocates":[89],"for":[90],"decision":[92],"measure":[94],"presence":[97],"shifts":[98,228],"population's":[100,263],"behavior.":[101,126],"find":[103],"alerts":[105],"judged":[106],"almost":[107],"identically":[108],"an":[110,266],"isolated":[111],"agent":[112],"produce":[114],"sharply":[115],"different":[116],"collective":[117,128],"outcomes.":[118,169],"Individual":[119],"audits":[120],"therefore":[121],"cannot":[122],"reliably":[123,184],"Yet":[127],"vulnerability":[129],"is":[130,158],"not":[131,183,190,255],"unpredictable:":[132],"using":[133],"only":[134],"benign,":[135],"adversary-free":[136],"behavior,":[137],"calibrate":[139],"response":[141],"function":[142],"forecasts":[144],"shift":[150],"under":[151],"future":[153],"attack,":[154],"before":[155,265],"attack":[157,242,267],"run.":[159],"next":[161],"examine":[162],"interaction":[164],"time":[166],"shape":[167],"these":[168],"Allowing":[170],"observe":[173],"reasoning":[176],"weaken":[178],"adversarial":[179,226],"influence,":[180],"but":[181],"does":[182,254],"prevent":[185],"capture.":[186],"However,":[187],"capture":[188],"need":[189],"permanent.":[192],"After":[193],"adversaries":[194],"are":[195,217],"removed":[196],"replaced,":[198],"mean":[199],"behavior":[201,264],"returns":[202],"toward":[203],"original":[205],"level,":[206],"even":[207],"after":[208],"five":[209],"thirty":[211],"rounds":[212],"pressure.":[214],"These":[215],"findings":[216],"consistent":[218],"with":[219],"calibrated":[221],"single-stable-state":[222],"model,":[223],"which":[225],"pressure":[227],"operating":[230],"point":[231],"without":[232],"creating":[233],"second":[235],"self-sustaining":[236],"state.":[237],"The":[238],"same":[239],"calibration":[240],"predicts":[241],"responses":[243],"24":[247],"48":[249],"Alignment":[251],"isolation":[253],"guarantee":[256],"alignment":[257],"population.":[260],"reveal":[269],"it":[271],"arrives.":[276]},"counts_by_year":[{"year":2026,"cited_by_count":2}],"updated_date":"2026-11-08T12:12:28.789561","created_date":"2026-08-26T00:00:00"}