[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-291167-105":59,"doc-detail-291167-en":131},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":124,"head_meta":126,"extra_data":128,"updated_unix":130},105,"en","trusted-machine-learning-models-unlock-private-inference-for-problems-currently-infeasible-with-cryptography","Trusted Machine Learning Models Unlock Private Inference for Problems Currently Infeasible with Cryptography","","Trusted Machine Learning Models Unlock Private Inference for Problems Currently Infeasible with Cryptography proposes Trusted Capable Model Environments (TCMEs) as an alternative to traditional cryptographic approaches for privacy-preserving interaction with untrusted parties. The paper argues that capable ML models can act as a trusted third party by operating under explicit input/output constraints, with explicit information-flow control and statelessness. It presents use cases enabled by TCME, shows that even some classic cryptographic problems can be solved, and discusses current limitations and implementation directions.",{"@graph":69,"@context":123},[70,84,106],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/trusted-machine-learning-models-unlock-private-inference-for-problems-currently-infeasible-with-cryptography/291167/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/trusted-machine-learning-models-unlock-private-inference-for-problems-currently-infeasible-with-cryptography/291167.png","ImageObject",300,407,{"name":92,"@type":93},"Aurelia","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-31","2026-09-17",true,{"@type":102,"interactionType":103,"userInteractionCount":105},"InteractionCounter",{"@type":104},"ViewAction",15,{"@type":107,"mainEntity":108},"FAQPage",[109,115,119],{"name":110,"@type":111,"acceptedAnswer":112},"What problem does TCME aim to solve?","Question",{"text":113,"@type":114},"TCME targets privacy-preserving computations between untrusted parties that require sharing private data in conventional settings. It enables secure private inference when cryptographic solutions are computationally infeasible due to the computation’s unstructured nature.","Answer",{"name":116,"@type":111,"acceptedAnswer":117},"What are the three fundamental properties required for models to be trusted?",{"text":118,"@type":114},"The paper lists statelessness, explicit information flow control (including frozen, verifiable information flow), and trustworthy and capable model(s) aligned with task expectations.",{"name":120,"@type":111,"acceptedAnswer":121},"How does TCME compare to classic cryptographic approaches like MPC and zero-knowledge proofs?",{"text":122,"@type":114},"The paper argues that TCMEs provide a new inference paradigm that can achieve secure private computation even when classical cryptographic methods are too costly for the applications considered. It also outlines a comparison with MPC and ZKPs and presents TCME-enabled solutions.","https://schema.org",{"og:url":83,"og:type":125,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":127,"canonical":83},"index,follow",{"doc_id":129,"site_id":62},291167,1789643585,{"code":4,"msg":5,"data":132},{"doc_id":129,"user_id":133,"nickname":92,"user_avatar":134,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":135,"file_id":136,"file_url":137,"file_type":138,"file_size":139,"view_count":105,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":140,"language":141,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":142,"faqs":143,"seo_title":144,"seo_description":67,"update_tm":130,"read_time":41,"preview_page_count":4},1099514068365,"https://ap-avatar.wpscdn.com/avatar/10000253d8d9f28188e?x-image-process=image/resize,m_fixed,w_180,h_180&k=1794309523707217728","arXiv :2501 .08970v1 [ cs .CR] 15 Jan 2025  \nTrusted Machine Learning Models Unlock Private Inference for Problems Currently Infeasible with Cryptography  \nIlia Shumailov1 , Daniel Ramage2 , Sarah Meiklejohn3 , Peter Kairouz2 , Florian Hartmann1 , Borja Balle1 and Eugene Bagdasarian2  \n* Inverse alphabetic order, 1 Google DeepMind, 2 Google Research, 3 Google  \nWe often interact with untrusted parties. Prioritization of privacy can limit the effectiveness of these interactions, as achieving certain goals necessitates sharing private data. Traditionally, addressing this challenge has involved either seeking trusted intermediaries or constructing cryptographic protocols that restrict how much data is revealed, such as multi-party computations or zero-knowledge proofs. While significant advances have been made in scaling cryptographic approaches, they remain limited in terms of the size and complexity of applications they can be used for. In this paper, we argue that capable machine learning models can fulfill the role of a trusted third party, thus enabling secure computations for applications that were previously infeasible. In particular, we describe Trusted Capable Model Environments (TCMEs) as an alternative approach for scaling secure computation, where capable machine learning model(s) interact under input/output constraints, with explicit information flow control and explicit statelessness. This approach aims to achieve a balance between privacy and computational efficiency, enabling private inference where classical cryptographic solutions are currently infeasible. We describe a number of use cases that are enabled by TCME, and show that even some simple classic cryptographic problems can already be solved with TCME. Finally, we outline current limitations and discuss the path forward in implementing them.  \n1. What are TCMEs?  \nIn this paper we contend that recent advancements in machine learning enable a new paradigm for private inference. Fundamentally, the need for many cryptographic primitives stems from the fact that we don’t have trusted third parties, thus requiring mutually untrusted participants to interact in a way that avoids revealing their data to each other but where they can nevertheless agree on a result. In this paper we argue that a capable machine learning model, in some settings, can play the role of a trusted third party (Abadi, 2004; Anderson, 2010) . We propose Trusted Capable Model Environments, a setting where an individual machine learning model or a number of models initiate an interaction with additional constraints in the input and output to ensure that private data cannot leave the TCME. Consider for example the classical millionaires problem, where a pair of individuals are trying to figure out who has more money without disclosing how much money they have. Cryptographically, this can be solved using a secure two-party computation by Yao, or subsequent protocols that are more efficient (Ioannidis and Grama, 2003; Lin and Tzeng, 2005; Yao, 1982) . With TCME, both individuals agree on  \n• A model, e.g., Gemma (Gemma-Team, 2024);  \n• A prompt, e.g.,“Say \"first\" if A is bigger than B and \"second\" otherwise? A={} and B={}.”;  \n• Input constraints, e.g., A and B are 32-bit integers;  \n• Output constraints, e.g., the only approved outputs are ‘first’ or ‘second’.  \nIf the environment can be trusted to both avoid leaking the private information provided by each  \nCorresponding author(s): [iliashumailov@google.com](iliashumailov@google.com)  \n© 2025 Google DeepMind. All rights reserved  \nparty and to reliably output the correct answer, this presents an alternate approach to enabling this computation. While a viable cryptographic solution exists for this simple example, for the applications we discuss below it is currently computationally infeasible to rely on cryptographic solutions due to the unstructured nature of the computation. In contrast, we argue that securely performing these computations ","cbCaieNdVEUnKTDv","https://ap.wps.com/l/cbCaieNdVEUnKTDv","pdf",683521,12,"English","# Trusted Capable Model Environments (TCMEs)\n## Why private inference needs trust and constraints\n## Three fundamental properties for trusted models\n## Comparison with cryptographic solutions (MPC and ZKPs)\n## TCME-enabled applications and limitations","[{\"question\":\"What problem does TCME aim to solve?\",\"answer\":\"TCME targets privacy-preserving computations between untrusted parties that require sharing private data in conventional settings. It enables secure private inference when cryptographic solutions are computationally infeasible due to the computation’s unstructured nature.\"},{\"question\":\"What are the three fundamental properties required for models to be trusted?\",\"answer\":\"The paper lists statelessness, explicit information flow control (including frozen, verifiable information flow), and trustworthy and capable model(s) aligned with task expectations.\"},{\"question\":\"How does TCME compare to classic cryptographic approaches like MPC and zero-knowledge proofs?\",\"answer\":\"The paper argues that TCMEs provide a new inference paradigm that can achieve secure private computation even when classical cryptographic methods are too costly for the applications considered. It also outlines a comparison with MPC and ZKPs and presents TCME-enabled solutions.\"}]","Trusted Machine Learning Models Unlock Private Inference for Problems Currently Infeasible with Cryptography | PDF"]