[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"$fh3EkcS1AdWb1X5C5Al96jvut_Bs8UI3qaNaO3AfJ9_Y":3},{"article":4,"iocs":54,"watch_terms":64,"solutions":65,"lesson":155},{"id":5,"title":6,"slug":7,"summary":8,"ai_summary":9,"brief":10,"url":11,"image_url":12,"published_at":13,"ingested_at":14,"relevance_score":15,"entities":16,"category_id":31,"category":32,"article_tags":36},"4a9c072f-c74c-45d9-a5f5-8dfea93197b2","OpenAI, Anthropic, Google API Flaw Let Weaker AI Models Decode Stronger Models' Reasoning","openai-anthropic-google-api-flaw-let-weaker-ai-models-decode-stronger-models-rea-685f5a","A newly disclosed flaw in the way OpenAI, Anthropic, and Google carried hidden AI reasoning between API calls let researchers recover internal reasoning and secrets from session logs, including API keys and passwords. The weakness affected encrypted reasoning objects used by the providers' reasoning APIs, where a block created in one session could be replayed into another and, during testing,","A vulnerability in the reasoning APIs of OpenAI, Anthropic, and Google allowed researchers to extract sensitive information, including API keys and passwords, from session logs. The flaw involved encrypted reasoning objects that could be replayed into different sessions or even handed to weaker models, effectively acting as a 'fuzzy' decoder. While the issue has been mitigated and is no longer reproducible, the researchers demonstrated potential for model distillation, data extraction, and hidden prompt injection.","Flaw in OpenAI, Anthropic, Google APIs allowed weaker models to decode stronger models' reasoning.","https:\u002F\u002Fthehackernews.com\u002F2026\u002F08\u002Fopenai-anthropic-google-api-flaw-let.html","https:\u002F\u002Fblogger.googleusercontent.com\u002Fimg\u002Fb\u002FR29vZ2xl\u002FAVvXsEiZP21lJn1EY_0JXWBul8gpBgRD_ryI4ACYqFiu6icbKCgIMyta9UqjQrtnzjTkO1Yi9tdzTaQw6X949lzMwtQPKUPPIxA5_EwFjJqjMDwBtFSc56vaCyybwJNcsjVZMfMW4KHQ_VVCjz3AVeovFEtsI8ENZTQmQuc9rDIF5yd0n9uciJmwqlWS_EKf5mo\u002Fs1600\u002Fai-models.jpg","2026-08-12T11:47:38+00:00","2026-08-12T14:00:16.090881+00:00",8,[17,20,22,24,26,28],{"name":18,"type":19},"OpenAI","vendor",{"name":21,"type":19},"Anthropic",{"name":23,"type":19},"Google",{"name":25,"type":19},"Microsoft",{"name":27,"type":19},"Hugging Face",{"name":29,"type":30},"Claude Haiku 4.5","product","80544778-fabb-4dcd-aa35-17492e5dcf4f",{"id":31,"icon":33,"name":34,"slug":35},null,"Vulnerabilities","vulnerabilities",[37,42,44,49],{"category":38},{"id":39,"icon":33,"name":40,"slug":41},"2e06f76c-d5b9-4f54-9eef-4d3447b10730","Breaches","breaches",{"category":43},{"id":31,"icon":33,"name":34,"slug":35},{"category":45},{"id":46,"icon":33,"name":47,"slug":48},"839da5c1-3c34-47e2-9499-f7201640e3ac","AI Security","ai-security",{"category":50},{"id":51,"icon":33,"name":52,"slug":53},"e7b231c8-5f79-4465-8d38-1ef13aea5a14","Threat Intelligence","threat-intelligence",[55,58,61],{"type":56,"value":29,"context":57},"malware","Used as a decoder model for Claude traces.",{"type":56,"value":59,"context":60},"GPT-5.6 Luna","Used as a decoder model for GPT traces.",{"type":56,"value":62,"context":63},"Gemini Robotics ER-1.6","Used as a decoder model for Gemini traces.",[23,25,18],{"controls":66,"vendors":149},[67,96,124],{"slug":68,"name":69,"problem":70,"products":71},"secrets-management","Secrets management","API keys, tokens and passwords stop leaking from code, chat and configuration files.",[72,80,88],{"vendor_slug":73,"vendor_name":74,"vendor_logo":75,"product_slug":76,"product_name":77,"url":78,"description":79},"hashicorp","HashiCorp","\u002Fvendor-logos\u002Fhashicorp.png","vault","HashiCorp Vault","https:\u002F\u002Fwww.hashicorp.com\u002Fen\u002Fproducts\u002Fvault","Secrets management with short-lived credentials and encryption as a service.",{"vendor_slug":81,"vendor_name":82,"vendor_logo":83,"product_slug":84,"product_name":85,"url":86,"description":87},"gitguardian","GitGuardian","\u002Fvendor-logos\u002Fgitguardian.png","gitguardian-platform","GitGuardian Secrets Detection","https:\u002F\u002Fwww.gitguardian.com\u002F","Finds and helps remediate secrets leaked in repositories, CI pipelines and chat tools.",{"vendor_slug":89,"vendor_name":90,"vendor_logo":91,"product_slug":92,"product_name":93,"url":94,"description":95},"palo-alto-networks","Palo Alto Networks","\u002Fvendor-logos\u002Fpalo-alto-networks.png","idira-secrets-management","Idira Secrets Management","https:\u002F\u002Fwww.paloaltonetworks.com\u002Fidira\u002Fmachine\u002Fsecrets-management","Secrets management for applications, scripts and pipelines, from the former CyberArk portfolio.",{"slug":97,"name":98,"problem":99,"products":100},"data-loss-prevention","Data loss prevention (DLP)","Sensitive data cannot quietly leave by email, upload, USB stick or AI tool.",[101,109,116],{"vendor_slug":102,"vendor_name":103,"vendor_logo":104,"product_slug":105,"product_name":106,"url":107,"description":108},"proofpoint","Proofpoint","\u002Fvendor-logos\u002Fproofpoint.png","enterprise-dlp","Proofpoint Enterprise DLP","https:\u002F\u002Fwww.proofpoint.com\u002Fus\u002Fproducts\u002Fdata-loss-prevention","Data loss prevention across email, cloud and endpoints.",{"vendor_slug":110,"vendor_name":25,"vendor_logo":111,"product_slug":112,"product_name":113,"url":114,"description":115},"microsoft","\u002Fvendor-logos\u002Fmicrosoft.png","purview-dlp","Microsoft Purview Data Loss Prevention","https:\u002F\u002Fwww.microsoft.com\u002Fen-us\u002Fsecurity\u002Fbusiness\u002Fdata-security-governance\u002Fmicrosoft-purview-data-security","Data classification and loss prevention across Microsoft 365, endpoints and connected cloud apps.",{"vendor_slug":117,"vendor_name":118,"vendor_logo":119,"product_slug":120,"product_name":121,"url":122,"description":123},"forcepoint","Forcepoint","\u002Fvendor-logos\u002Fforcepoint.png","forcepoint-dlp","Forcepoint DLP","https:\u002F\u002Fwww.forcepoint.com\u002Fproduct\u002Fdlp-data-loss-prevention","Data loss prevention for endpoints, network, email and cloud applications.",{"slug":48,"name":125,"problem":126,"products":127},"AI application security","Prompt injection, data leaks through AI tools and unapproved AI use get caught.",[128,136,144],{"vendor_slug":129,"vendor_name":130,"vendor_logo":131,"product_slug":132,"product_name":133,"url":134,"description":135},"cisco","Cisco","\u002Fvendor-logos\u002Fcisco.png","ai-defense","Cisco AI Defense","https:\u002F\u002Fwww.cisco.com\u002Fsite\u002Fus\u002Fen\u002Fproducts\u002Fsecurity\u002Fai-defense\u002Findex.html","Discovery of AI use, testing of models and guardrails for AI applications.",{"vendor_slug":137,"vendor_name":138,"vendor_logo":139,"product_slug":140,"product_name":141,"url":142,"description":143},"hiddenlayer","HiddenLayer","\u002Fvendor-logos\u002Fhiddenlayer.png","hiddenlayer-ai-security","HiddenLayer AI Security Platform","https:\u002F\u002Fhiddenlayer.com\u002F","Model scanning, AI red teaming and runtime protection for AI applications.",{"vendor_slug":89,"vendor_name":90,"vendor_logo":91,"product_slug":145,"product_name":146,"url":147,"description":148},"prisma-airs","Prisma AIRS","https:\u002F\u002Fwww.paloaltonetworks.com\u002Fai-security\u002Fprisma-airs","Security for AI applications, models and agents, including runtime protection against prompt injection.",[150],{"slug":151,"name":23,"logo_path":152,"roles":153},"google","\u002Fvendor-logos\u002Fgoogle.png",[154],"affected",{"slug":156,"title":157,"body":158,"framework_refs":159,"framework_notes":171},"ai-reasoning-apis-leaked-secrets-via-replayable-encrypted-objects","AI Reasoning APIs Leaked Secrets via Replayable Encrypted Objects","A design flaw in the reasoning APIs of OpenAI, Anthropic, and Google allowed encrypted reasoning objects to be replayed across different sessions or fed to weaker AI models, inadvertently exposing sensitive data such as API keys and passwords. The root issue was insufficient session binding and context isolation — encrypted objects were treated as portable and reusable rather than tightly scoped to their originating session. This matters because it demonstrates that encryption alone is not a sufficient security control; proper session integrity, token binding, and output sanitization are equally critical. The flaw also opens the door to model distillation attacks and hidden prompt injection, representing a novel class of AI-specific threats that traditional security frameworks have not yet fully addressed.",[160,161,162,163,164,165,166,167,168,169,170],"NIST SP 800-53 AC-12 (Session Termination)","NIST SP 800-53 SC-23 (Session Authenticity)","NIST SP 800-53 SI-10 (Information Input Validation)","NIST SP 800-53 IA-11 (Re-Authentication)","CIS Control 3: Data Protection","CIS Control 16: Application Software Security","OWASP API Security Top 10 - API2: Broken Authentication","OWASP API Security Top 10 - API8: Security Misconfiguration","OWASP ML Security Top 10 - ML05: Model Inversion Attack","GDPR Article 32 (Security of Processing)","NIST AI RMF - Govern 1.7 (AI Risk Policies)",{}]