[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-82275-en":3,"doc-seo-82275-105":29,"detail-sidebar-cat-0-en-105":91},{"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":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},82275,962075114101,"Seraphina","https://ap-avatar.wpscdn.com/avatar/e000253a75eb197efd?x-image-process=image/resize,m_fixed,w_180,h_180&k=1780044092746381165",8,"Research & Report","Geopolitical Alignment Endorsement Effects in Large Language Models","Large language models (LLMs) are increasingly used to summarize and evaluate policy-relevant information, yet whether their judgments are shaped by geopolitical cues remains unclear. An endorsement experiment tests four LLMs evaluating the same international economic and security policies while the endorsing actor is randomized among the United States, European Union, China, and Russia. Models show endorsement-dependent approval scores and explanation-driven shifts, indicating that endorser identity can modulate perceived credibility, security, sovereignty, surveillance, and geopolitical risk even with fixed policy content.","arXiv :2607 .09262v1 [ cs .CY] 10 Jul 2026  \nGeopolitical alignment: Endorsement effects in large language models  \nMaxim Chupilkin∗  \nAbstract  \nLarge language models (LLMs) are increasingly used to summarize and evaluate policy-relevant information, but it remains unclear whether their judgments are implicitly shaped by geopolitical cues. I study this question with an endorsement experiment in which four LLMs evaluate the same international economic and security policies after each policy is randomly described as supported by the United States, the European Union, China, or Russia. In the numeric-only condition, GPT-5, Claude Sonnet, and Gemini rate China-and Russia-endorsed policies substantially lower than identical policies endorsed by the United States or the European Union; DeepSeek is the main exception. A second condition asks models to provide a short justification with the score. This request leaves the broad Western/non-Western gap intact for GPT-5 and Claude Sonnet, attenuates Gemini’s penalties, and sharply activates China and Russia penalties in DeepSeek. The justifications indicate that Western endorsement is often treated as a credibility cue, whereas Chinese and Russian endorsement is treated as a cue for data security, sovereignty, surveillance, or geopolitical risk. These findings show that LLM policy evaluations can depend on the identity of a foreign endorser even when policy content is held fixed.  \nKeywords: large language models; artificial intelligence; geopolitical bias; policy evaluation; endorsement experiments  \nIntroduction  \nLarge language models are becoming intermediaries in policy work. They summarize briefings, draft assessments, compare policy designs, and increasingly produce judgments that may inform human decisions. Because foundation models are trained and adapted at broad scale, their downstream use can carry sociotechnical risks that are difficult to infer from benchmark performance alone [1–4] . These uses raise the question of whether models carry implicit geopolitical leanings that shape their evaluations even when policy content is held fixed. This question is especially salient  \n∗ Department of Politics and International Relations, University of Oxford. Email: [maxim.chupilkin@politics.ox.ac.uk](maxim.chupilkin@politics.ox.ac.uk)  \nas governments increasingly frame AI capability, infrastructure, and governance in terms of sovereignty and strategic competition, including competition between the United States and China.  \nI test this possibility with an endorsement experiment. The design holds policy content fixed and randomizes only the foreign actor described as supporting the policy. The endorsers are the United States, the European Union, China, and Russia. The policies are moderate and technocratic: one concerns a shared digital customs platform, and the other concerns a shared cyber incident reporting platform. The outcome is the model’s approval score on a 0–100 scale. I apply the design to OpenAI GPT-5, Anthropic Claude Sonnet, Google Gemini 2.5 Flash, and DeepSeek Chat, and then repeat the experiment with a prompt that requires a short justification. This second condition tests whether explanation requests merely reveal the basis of evaluation or also alter the evaluation itself.  \nThe results show clear geopolitical endorsement effects. In the numeric-only condition, GPT-5, Claude Sonnet, and Gemini score China- and Russia-endorsed policies lower than policies endorsed by the United States or the European Union. The pattern differs by model. Claude Sonnet is most sensitive to Chinese and Russian endorsement in the security vignette. Gemini applies large penalties even in the economic vignette. DeepSeek is the baseline exception: it shows little systematic differentiation across endorsers when asked only for a number. Justification prompts are not merely diagnostic: they can also change the evaluation being measured, consistent with evidence that generated explanations ca","cbCaiq5C2TvOOsoS","https://ap.wps.com/l/cbCaiq5C2TvOOsoS","pdf",1298714,1,20,"English","en",105,"# Abstract\n# Introduction\n## Experimental design and conditions\n## Results and model differences\n## Related work and motivation","[{\"question\":\"How does the endorsement experiment isolate geopolitical effects on LLM policy judgments?\",\"answer\":\"The experiment holds policy content fixed and randomizes only the foreign endorser described as supporting the policy, across the United States, European Union, China, and Russia.\"},{\"question\":\"What pattern of approval scores appears in the numeric-only condition?\",\"answer\":\"GPT-5, Claude Sonnet, and Gemini generally rate China- and Russia-endorsed policies lower than those endorsed by the United States or the European Union, while DeepSeek shows little systematic differentiation.\"},{\"question\":\"How do justification prompts change the evaluations?\",\"answer\":\"Requesting a short justification affects measured evaluations, not just diagnostics: it leaves the broad Western/non-Western gap intact for GPT-5 and Claude Sonnet, attenuates Gemini’s penalties, and sharply activates China and Russia penalties in DeepSeek.\"}]",1784179332,50,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"geopolitical-alignment-endorsement-effects-in-large-language-models","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/geopolitical-alignment-endorsement-effects-in-large-language-models/82275/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-22","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"How does the endorsement experiment isolate geopolitical effects on LLM policy judgments?","Question",{"text":75,"@type":76},"The experiment holds policy content fixed and randomizes only the foreign endorser described as supporting the policy, across the United States, European Union, China, and Russia.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What pattern of approval scores appears in the numeric-only condition?",{"text":80,"@type":76},"GPT-5, Claude Sonnet, and Gemini generally rate China- and Russia-endorsed policies lower than those endorsed by the United States or the European Union, while DeepSeek shows little systematic differentiation.",{"name":82,"@type":73,"acceptedAnswer":83},"How do justification prompts change the evaluations?",{"text":84,"@type":76},"Requesting a short justification affects measured evaluations, not just diagnostics: it leaves the broad Western/non-Western gap intact for GPT-5 and Claude Sonnet, attenuates Gemini’s penalties, and sharply activates China and Russia penalties in DeepSeek.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,114,119,122,126,129,133],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":28,"slug":113},6,"Technology","technology",{"id":115,"doc_module":4,"doc_module_name":45,"category_name":116,"show_sort_weight":117,"slug":118},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":45,"category_name":124,"show_sort_weight":21,"slug":125},9,"Religion & Spirituality","religion-spirituality",{"id":21,"doc_module":4,"doc_module_name":45,"category_name":127,"show_sort_weight":21,"slug":128},"World Cup","world-cup",{"id":130,"doc_module":4,"doc_module_name":45,"category_name":131,"show_sort_weight":130,"slug":132},10,"Lifestyle","lifestyle",{"id":134,"doc_module":4,"doc_module_name":45,"category_name":135,"show_sort_weight":106,"slug":136},19,"General","general"]