74
likes
13
comments
daily view 0
monthly view 0
Live Analytics
Comments 0
This API Flaw "Leaked" Passwords From AI's Hidden Reasoning Analytics Table
Income Estimates for This API Flaw "Leaked" Passwords From AI's Hidden Reasoning
Based on this YouTube video's total view count of 3.83K views and industry-standard rates, the estimated total earning is $3 - $8 through ad revenue. Historical data is not yet available to calculate daily, weekly, or monthly averages.
About This API Flaw "Leaked" Passwords From AI's Hidden Reasoning
Explore This API Flaw "Leaked" Passwords From AI's Hidden Reasoning with 3,831 views, 74 likes, and 13 comments. Experience the impact of this video content that has captured audience attention.
Why did Anthropic, OpenAI, and Google's "encrypted" chain-of-thought reasoning blocks leak private API keys, passwords, and system tokens when passed to cheaper models? Meet the architectural flaw in client-side stateful AEAD envelopes and the 315,000 decoded reasoning blocks exposing developer credentials. 🔔 Subscribe: https://www.youtube.com/channel/UC0DZj1PNa_Fp0MU6uPSKv5w?sub_confirmation=1 💙 Become a Member: https://www.youtube.com/channel/UC0DZj1PNa_Fp0MU6uPSKv5w/join 🐦 Twitter/X: https://x.com/cloud_codes 💬 Discord: https://discord.gg/4kJqEBMMf In this deep dive, Cloud Codes breaks down the system design, cryptographic mechanics, and security vulnerabilities of client-side LLM reasoning envelopes. We examine how ELLIS Institute researchers (Panfilov et al.) used cheap sibling models like Haiku 4.5 and Luna to transcribe encrypted thinking traces in plaintext, proving extraction accuracy by matching generated token lengths directly against billed API invoices. Furthermore, we analyze 6,708 public agent trajectories on GitHub and HuggingFace containing 62 leaked API keys and 33 passwords, demonstrate cross-account signature replay attacks and prompt injection via reasoning blocks, review Matthew Green's ignored cryptographer bug bounty reports, and explain why moving state back to server-side session handles is the only permanent architectural fix. If this helped you understand backend architecture, system design, and how to build faster software, subscribe to Cloud Codes for a new infrastructure breakdown every single week! Build, solve, deploy. 🔗 Repositories & Sources Mentioned: • Panfilov et al. Paper (arXiv:2608.09867): https://arxiv.org/abs/2608.09867 • Matthew Green's Encrypted Reasoning Post: https://blog.cryptographyengineering.com/2026/05/29/fooling-around-with-encrypted-reasoning-blobs/ • Simon Willison AI Security Analysis: https://simonwillison.net/ • Anthropic Claude API Extended Thinking Docs: https://docs.anthropic.com/ • OpenAI Reasoning & o1 Architecture Overview: https://platform.openai.com/docs/guides/reasoning ⏱️ Video Chapters: 0:00 - The Base64 Encrypted Thinking Block 1:11 - Why Chain of Thought Went Behind a Curtain 2:30 - AEAD Envelopes & The Stateless API Design 3:24 - The Attack: Replaying Signatures into Cheaper Models 5:35 - Verifying Extraction: Billed Tokens vs Output Length 6:30 - Scanning 6,708 Public Trajectories for Leaked Credentials 8:44 - Bypassing Safety Refusals & Prompt Injection Attacks 10:40 - Matthew Green's Ignored Bug Bounty Reports 12:24 - Capability Stealing & The Anti-Distillation Moat 14:17 - Output Token Billing vs Security Mitigations 17:37 - Final Verdict: Server-Side State & Sanitizing Trajectories #aisecurity #claude #openai #cybersecurity #systemdesign #cloudcodes #anthropic #backend #devops #softwareengineering User Queries: ai api encrypted reasoning leak panfilov ellis institute claude opus thinking block signature extraction haiku openai aead envelope cross account replay attack how reasoning block leaks api keys in public trajectories matthew green cryptographer llm reasoning replay bug bounty prompt injection inside encrypted chain of thought reasoning server side state vs stateless client reasoning envelopes chain of thought distillation security risks llm huggingface public agent trajectory credential exposure cloud codes ai api flaw leaked passwords breakdown
About YouTube Real-Time View Count
With SocialCounts.org’s view counter, track your YouTube video’s live view count and YouTube likes count in real time with fast, reliable updates.
Watch every YouTube video live view count rise with our real-time YouTube views tracker—built for accuracy and minimal delay.
Follow YouTube real time views as they happen, using our dedicated view counter for YouTube videos.
Get up-to-date live view count on YouTube and see real-time growth with SocialCounts.org’s smart tracking tools.
Embed Widget
Parameters:
fullscreen=true- Fullscreen countergraph=true- Live graph chartcounter=0/1/2- Select counter (0=likes, 1=views, 2=comments)
URL
Click to copy the embed URL to your clipboard

