{"id":"https://openalex.org/W7154304614","doi":"https://doi.org/10.48550/arxiv.2604.10577","title":"The Blind Spot of Agent Safety: How Benign User Instructions Expose Critical Vulnerabilities in Computer-Use Agents","display_name":"The Blind Spot of Agent Safety: How Benign User Instructions Expose Critical Vulnerabilities in Computer-Use Agents","publication_year":2026,"publication_date":"2026-04-12","ids":{"openalex":"https://openalex.org/W7154304614","doi":"https://doi.org/10.48550/arxiv.2604.10577"},"language":null,"primary_location":{"id":"doi:10.48550/arxiv.2604.10577","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.10577","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,"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"},"type":"preprint","indexed_in":["datacite"],"open_access":{"is_oa":true,"oa_status":"green","oa_url":"https://doi.org/10.48550/arxiv.2604.10577","any_repository_has_fulltext":true},"authorships":[{"author_position":"first","author":{"id":"https://openalex.org/A5133589996","display_name":"Xuwei Ding","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Ding, Xuwei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133620493","display_name":"Skylar Zhai","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhai, Skylar","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133610820","display_name":"Linxin Song","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Song, Linxin","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133621422","display_name":"Jiate Li","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Li, Jiate","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133564817","display_name":"Taiwei Shi","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Shi, Taiwei","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5058072829","display_name":"Nicholas Meade","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Meade, Nicholas","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5108816545","display_name":"Siva Reddy","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Reddy, Siva","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"middle","author":{"id":"https://openalex.org/A5133569318","display_name":"Jian Kang","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Kang, Jian","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]},{"author_position":"last","author":{"id":"https://openalex.org/A5133612402","display_name":"Jieyu Zhao","orcid":null},"institutions":[],"countries":[],"is_corresponding":false,"raw_author_name":"Zhao, Jieyu","raw_affiliation_strings":[],"raw_orcid":null,"affiliations":[]}],"institutions":[],"countries_distinct_count":0,"institutions_distinct_count":0,"corresponding_author_ids":[],"corresponding_institution_ids":[],"apc_list":null,"apc_paid":null,"fwci":null,"has_fulltext":false,"cited_by_count":0,"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/T11424","display_name":"Security and Verification in Computing","score":0.37459999322891235,"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/T11424","display_name":"Security and Verification in Computing","score":0.37459999322891235,"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/T11689","display_name":"Adversarial Robustness in Machine Learning","score":0.20430000126361847,"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/T10734","display_name":"Information and Cyber Security","score":0.06800000369548798,"subfield":{"id":"https://openalex.org/subfields/1710","display_name":"Information Systems"},"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/context","display_name":"Context (archaeology)","score":0.6428999900817871},{"id":"https://openalex.org/keywords/harm","display_name":"Harm","score":0.5921000242233276},{"id":"https://openalex.org/keywords/vulnerability","display_name":"Vulnerability (computing)","score":0.5859000086784363},{"id":"https://openalex.org/keywords/task","display_name":"Task (project management)","score":0.5551999807357788},{"id":"https://openalex.org/keywords/expose","display_name":"EXPOSE","score":0.36090001463890076},{"id":"https://openalex.org/keywords/unintended-consequences","display_name":"Unintended consequences","score":0.31220000982284546}],"concepts":[{"id":"https://openalex.org/C41008148","wikidata":"https://www.wikidata.org/wiki/Q21198","display_name":"Computer science","level":0,"score":0.7253999710083008},{"id":"https://openalex.org/C38652104","wikidata":"https://www.wikidata.org/wiki/Q3510521","display_name":"Computer security","level":1,"score":0.6456000208854675},{"id":"https://openalex.org/C2779343474","wikidata":"https://www.wikidata.org/wiki/Q3109175","display_name":"Context (archaeology)","level":2,"score":0.6428999900817871},{"id":"https://openalex.org/C2777363581","wikidata":"https://www.wikidata.org/wiki/Q15098235","display_name":"Harm","level":2,"score":0.5921000242233276},{"id":"https://openalex.org/C95713431","wikidata":"https://www.wikidata.org/wiki/Q631425","display_name":"Vulnerability (computing)","level":2,"score":0.5859000086784363},{"id":"https://openalex.org/C2780451532","wikidata":"https://www.wikidata.org/wiki/Q759676","display_name":"Task (project management)","level":2,"score":0.5551999807357788},{"id":"https://openalex.org/C108827166","wikidata":"https://www.wikidata.org/wiki/Q175975","display_name":"Internet privacy","level":1,"score":0.48350000381469727},{"id":"https://openalex.org/C107457646","wikidata":"https://www.wikidata.org/wiki/Q207434","display_name":"Human\u2013computer interaction","level":1,"score":0.421099990606308},{"id":"https://openalex.org/C2779137570","wikidata":"https://www.wikidata.org/wiki/Q16243196","display_name":"EXPOSE","level":2,"score":0.36090001463890076},{"id":"https://openalex.org/C2776889888","wikidata":"https://www.wikidata.org/wiki/Q1135789","display_name":"Unintended consequences","level":2,"score":0.31220000982284546},{"id":"https://openalex.org/C140547941","wikidata":"https://www.wikidata.org/wiki/Q7797194","display_name":"Threat model","level":2,"score":0.29330000281333923},{"id":"https://openalex.org/C2780522230","wikidata":"https://www.wikidata.org/wiki/Q1140419","display_name":"Ambiguity","level":2,"score":0.2915000021457672},{"id":"https://openalex.org/C14036430","wikidata":"https://www.wikidata.org/wiki/Q3736076","display_name":"Function (biology)","level":2,"score":0.2890999913215637},{"id":"https://openalex.org/C165696696","wikidata":"https://www.wikidata.org/wiki/Q11287","display_name":"Exploit","level":2,"score":0.2752000093460083},{"id":"https://openalex.org/C112930515","wikidata":"https://www.wikidata.org/wiki/Q4389547","display_name":"Risk analysis (engineering)","level":1,"score":0.2599000036716461}],"mesh":[],"locations_count":1,"locations":[{"id":"doi:10.48550/arxiv.2604.10577","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.10577","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,"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":null,"raw_source_name":null,"raw_type":"Preprint"}],"best_oa_location":{"id":"doi:10.48550/arxiv.2604.10577","is_oa":true,"landing_page_url":"https://doi.org/10.48550/arxiv.2604.10577","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,"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"},"sustainable_development_goals":[{"display_name":"Peace, Justice and strong institutions","id":"https://metadata.un.org/sdg/16","score":0.6516935229301453}],"awards":[],"funders":[],"has_content":{"grobid_xml":false,"pdf":false},"content_urls":null,"referenced_works_count":0,"referenced_works":[],"related_works":[],"abstract_inverted_index":{"Computer-use":[0],"agents":[1],"(CUAs)":[2],"can":[3,17],"now":[4],"autonomously":[5],"complete":[6],"complex":[7],"tasks":[8,77],"in":[9,141],"real":[10],"digital":[11],"environments,":[12],"but":[13,39],"when":[14,135,155],"misled,":[15],"they":[16],"also":[18],"be":[19],"used":[20],"to":[21,133,190,197,203],"automate":[22],"harmful":[23,182],"actions":[24],"programmatically.":[25],"Existing":[26],"safety":[27,150,209],"evaluations":[28],"largely":[29],"target":[30],"explicit":[31],"threats":[32,88],"such":[33],"as":[34],"misuse":[35],"and":[36,52,83,89,97,110,169,206],"prompt":[37],"injection,":[38],"overlook":[40],"a":[41,65],"subtle":[42],"yet":[43],"critical":[44],"setting":[45],"where":[46],"user":[47,156],"instructions":[48,157],"are":[49,158],"entirely":[50],"benign":[51],"harm":[53],"arises":[54],"from":[55,131,184],"the":[56,112,165,181,185,199],"task":[57],"context":[58],"or":[59],"execution":[60],"outcome.":[61],"We":[62,192],"introduce":[63],"OS-BLIND,":[64],"benchmark":[66],"that":[67,101,148],"evaluates":[68],"CUAs":[69,103],"under":[70],"unintended":[71],"attack":[72,106],"conditions,":[73],"comprising":[74],"300":[75],"human-crafted":[76],"across":[78],"12":[79],"categories,":[80],"8":[81],"applications,":[82],"2":[84],"threat":[85],"clusters:":[86],"environment-embedded":[87],"agent-initiated":[90],"harms.":[91],"Our":[92,144],"evaluation":[93],"on":[94],"frontier":[95],"models":[96,189],"agentic":[98],"frameworks":[99],"reveals":[100],"most":[102],"exceed":[104],"90%":[105],"success":[107],"rate":[108],"(ASR),":[109],"even":[111,125],"safety-aligned":[113,188],"Claude":[114,136],"4.5":[115,137],"Sonnet":[116,138],"reaches":[117],"73.0%":[118,132],"ASR.":[119],"More":[120],"interestingly,":[121],"this":[122],"vulnerability":[123],"becomes":[124],"more":[126],"severe,":[127],"with":[128],"ASR":[129],"rising":[130],"92.7%":[134],"is":[139],"deployed":[140],"multi-agent":[142,176],"systems.":[143],"analysis":[145],"further":[146,204],"shows":[147],"existing":[149],"defenses":[151],"provide":[152],"limited":[153],"protection":[154],"benign.":[159],"Safety":[160],"alignment":[161],"primarily":[162],"activates":[163],"within":[164],"first":[166],"few":[167],"steps":[168],"rarely":[170],"re-engages":[171],"during":[172],"subsequent":[173],"execution.":[174],"In":[175],"systems,":[177],"decomposed":[178],"subtasks":[179],"obscure":[180],"intent":[183],"model,":[186],"causing":[187],"fail.":[191],"will":[193],"release":[194],"our":[195],"OS-BLIND":[196],"encourage":[198],"broader":[200],"research":[201],"community":[202],"investigate":[205],"address":[207],"these":[208],"challenges.":[210]},"counts_by_year":[],"updated_date":"2026-07-28T07:46:37.118299","created_date":"2026-04-15T00:00:00"}