[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84992-en":3,"doc-seo-84992-105":30,"detail-sidebar-cat-0-en-105":90},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},84992,7971461740909,"Levi","https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d",8,"Research & Report","Initiation Safety A Missing Dimension in Generalist Robot Safety","Safety for generalist robots is often discussed as motion control or conversational behavior, leaving a critical gap: initiation authorization—whether the robot should take its first hard-to-undo social action when consent is unclear. The work argues that common stacks treat engagement scores or VLA rollout confidence as permission, conflating seeing a person with having their approval. It frames initiation authorization as a distinct safety layer, contrasts PAS with direct initiation on logged traces, and outlines a user study focusing on metrics, governance, and boundary conditions between initiation and foundation-model generation.","Initiation Safety: A Missing Dimension in Generalist-Robot Safety  \nZhijin Meng and Francisco Cruz  \nSchool of Computer Science and Engineering  \nUniversity of New South Wales  \nEmail: [zhijin.meng@unsw.edu.au](zhijin.meng@unsw.edu.au), [f.cruz@unsw.edu.au](f.cruz@unsw.edu.au)  \narXiv :2607 .07420v 1 [ cs .RO] 8 Jul 2026  \nAbstract—Safety for generalist robots is usually discussed in terms of motion or dialogue. We argue a third question is missing: should the robot take its first hard-to-undo social action at all, such as a greeting, an uninvited grasp, or stepping into someone’s space? We call this initiation authorization. Current frameworks rarely treat it as a separate safety layer. Today’s stacks often skip this step: a high engagement score or a confident VLA rollout is treated as permission to act. But seeing a person is not the same as having their consent to be addressed. We frame initiation authorization within generalist-robot safety and contrast it with post-plan VLA guardrails, implementing PAS (probe–authorize– speak) on a doorway humanoid, comparing it with direct-init on logged traces, and proposing a three-condition user study, with open questions on metrics, governance, and where initiation ends and foundation-model generation begins.  \nI. INTRODUCTION  \nGeneralist-robot safety must include initiation authorization: deciding whether a robot may take its first hard-toundo social action when the situation is still unclear. An engagement score measures how receptive someone seems. It does not answer whether the robot should act when that signal is ambiguous [9, 14, 11] .  \nDeployed stacks often distinguish two safety layers (physical and social-interaction) but treat initiation as part of the latter rather than as a separate decision. Physical safety limits motion (collisions, force, joint limits [15]) . Social-interaction safety governs how the robot behaves once interaction has already started, including recovery when a first act misfires [1, 13, 6] . We argue for a third layer that this picture largely omits: initiation authorization, whether the robot may start at all. The harm often looks like an ordinary first move: amistimed greeting that interrupts a call [4], or an end-toend VLA that reaches for an object before the person has agreed [7] . A jailbroken planner can authorize a harmful first move as readily as a harmful first word [10] .  \nMost deployed stacks treat initiation as a fixed threshold: speak ⇔ et > θ [3] . That treats an engagement score, or a single VLA rollout, as permission to act, with no context, no user preference, and no evidence from low-cost probes [12, 3] . Figure 1 contrasts this passive threshold with PAS (probe– authorize–speak): small reversible probes before any hard-toundo action, an explicit bold/conservative setting, and runtime checks before foundation-model dialogue.  \nBuilding on this distinction, we make three contributions:  \n1) We introduce initiation authorization as a distinct safety layer, separate from physical shields, post-plan VLA guardrails, post-dialogue preference alignment, and engagement optimization.  \n2) We illustrate PAS (probe–authorize–speak) on a physical robot and propose an evaluation protocol that would compare three initiation policies.  \n3) We elaborate on future work and pose open questions on metrics, governance, and where initiation ends and foundation-model generation begins.  \nII. INITIATION AUTHORIZATION IN GENERALIST-ROBOT SAFETY  \nA. Where Initiation Fits  \nPhysical safety (collisions, forces, joint limits) remains necessary [1, 15] . HRI already separates physical safety from how safe people feel [1], and social-error taxonomies list what goes wrong once interaction is already underway [13] . Less covered is whether the robot should start in the first place [11, 4] . We call that initiation authorization. Table I summarizes the three safety layers for generalist robots.  \nTABLE I: Safety layers for generalist robots.  \n\n| Layer | What i","cbCaipG4rSBy0g79","https://ap.wps.com/l/cbCaipG4rSBy0g79","pdf",226294,2,1,4,"English","en",105,"# Introduction\n## Initiation authorization as a third safety layer\n# Initiation Authorization in Generalist-Robot Safety\n## Where initiation fits\n## Passive thresholds vs PAS","[{\"question\":\"What is initiation authorization in generalist-robot safety?\",\"answer\":\"Initiation authorization decides whether a generalist robot may take its first hard-to-undo social action when the situation is still unclear.\"},{\"question\":\"Why are engagement scores or VLA confidence not sufficient for deciding to act first?\",\"answer\":\"They indicate receptiveness or system confidence, but do not establish consent for being addressed or acted upon, especially under ambiguous cues.\"},{\"question\":\"How does PAS (probe–authorize–speak) differ from a passive threshold policy?\",\"answer\":\"PAS uses small reversible probes without speech, collects evidence, and opens an explicit authorization gate before speaking or acting, whereas passive thresholds treat a score or rollout as immediate permission.\"}]",1784200090,10,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":85,"head_meta":87,"extra_data":89,"updated_unix":28},"initiation-safety-a-missing-dimension-in-generalist-robot-safety","",{"@graph":36,"@context":84},[37,52,67],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,47,50],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":20},"https://docshare.wps.com/document/","Document",{"item":48,"name":12,"@type":43,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":43,"position":22},"https://docshare.wps.com/document/initiation-safety-a-missing-dimension-in-generalist-robot-safety/84992/",{"url":51,"name":13,"@type":53,"author":54,"headline":13,"publisher":56,"fileFormat":59,"inLanguage":24,"description":14,"dateModified":60,"datePublished":61,"encodingFormat":59,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":55},"Person",{"url":41,"name":57,"@type":58},"DocShare","Organization","application/pdf","2026-07-23","2026-07-16",true,{"@type":64,"interactionType":65,"userInteractionCount":20},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70,76,80],{"name":71,"@type":72,"acceptedAnswer":73},"What is initiation authorization in generalist-robot safety?","Question",{"text":74,"@type":75},"Initiation authorization decides whether a generalist robot may take its first hard-to-undo social action when the situation is still unclear.","Answer",{"name":77,"@type":72,"acceptedAnswer":78},"Why are engagement scores or VLA confidence not sufficient for deciding to act first?",{"text":79,"@type":75},"They indicate receptiveness or system confidence, but do not establish consent for being addressed or acted upon, especially under ambiguous cues.",{"name":81,"@type":72,"acceptedAnswer":82},"How does PAS (probe–authorize–speak) differ from a passive threshold policy?",{"text":83,"@type":75},"PAS uses small reversible probes without speech, collects evidence, and opens an explicit authorization gate before speaking or acting, whereas passive thresholds treat a score or rollout as immediate permission.","https://schema.org",{"og:url":51,"og:type":86,"og:title":13,"og:site_name":57,"og:description":14},"article",{"robots":88,"canonical":51},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":91},[92,96,100,104,109,114,119,122,127,130,133],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":93,"show_sort_weight":94,"slug":95},"Story & Novel",90,"story-novel",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":97,"show_sort_weight":98,"slug":99},"Literature",80,"literature",{"id":22,"doc_module":4,"doc_module_name":46,"category_name":101,"show_sort_weight":102,"slug":103},"Exam",70,"exam",{"id":105,"doc_module":4,"doc_module_name":46,"category_name":106,"show_sort_weight":107,"slug":108},5,"Comic",60,"comic",{"id":110,"doc_module":4,"doc_module_name":46,"category_name":111,"show_sort_weight":112,"slug":113},6,"Technology",50,"technology",{"id":115,"doc_module":4,"doc_module_name":46,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":29,"doc_module":4,"doc_module_name":46,"category_name":131,"show_sort_weight":29,"slug":132},"Lifestyle","lifestyle",{"id":134,"doc_module":4,"doc_module_name":46,"category_name":135,"show_sort_weight":105,"slug":136},19,"General","general"]