Introduction to Chainlink Confidential AI Attester
summary
Alpha Chainlink Confidential AI Attester performs document analysis inside a Trusted Execution Environment and returns verifiable attestations that a DeFi contract can read on-chain. The demo uploads a mock financial PDF, runs an LLM (Gemma 4 or switchable models) inside AWS Nitro Enclave, receives an attestation that the borrower is approved, and a CRE callback decrypts and writes the decision, document digest, and transaction hash to testnet using CRE CLI v19+ --broadcast. Hackathon bounties include specific cash prizes for CRE, other Chainlink services, the attester, and continuity tracks.
AI-written summary · about
chapters (35)
- 0:00Introduction
- 0:12Why Chainlink for Your Hackathon Project
- 0:36Meet Solange and Andrej
- 1:00The Chainlink Bounties
- 1:31The New Confidential AI Attester Bounty
- 2:01Extending an Existing Chainlink Project
- 2:12Installing the Chainlink Developer Skills
- 2:29The Chainlink Runtime Environment
- 2:56A Special Guest Judge: Ari Juels
- 3:18What Is the Confidential AI Attester
- 3:42Why Proving Financials Is Not Private
- 4:02An Early Alpha Experiment
- 4:25Candidate Use Cases
- 4:48A Look at the Playground and Resources
- 5:33Why to Use the AI Skills
- 5:52Getting an API Key
- 6:16Uploading a Financial PDF
- 6:45Running the Analysis Inside a TEE
- 7:16Verifying the Attestation On Chain
- 7:49How the CRE Callback Stores the Decision
- 8:20The Undercollateralized Loan Prompt
- 8:40Checking the CRE CLI Version
- 9:05Simulating With an HTTP Trigger
- 9:29Exposing Localhost With ngrok
- 9:54Submitting the Request
- 10:18The Broadcast Flag Writes to Testnet
- 11:07Seeing the Approved Decision and Transaction
- 11:34Querying the Contract for the Decision
- 12:02The LLM and Enclave Behind It
- 12:36Q&A: Is CRE Required
- 13:04Walking Through the Demo Repo
- 14:07Q&A: Verifying the Callback
- 15:21Recapping the Four Bounties
- 16:35API Keys and Data Streams Credentials
- 17:15Closing
related talks
02h 06mDeep Dive Into Functions
41mAgentic AI, confidential computing and attested TLS
01h 59mLLM applications with LangChain
25mLangChain with Azure AI: Building LLM powered Applications
46mBeyond Black-Box AI: Designing Transparent Automation for Regulated Industries
27mFrom Perimeter to Silicon: Why Agentic AI Demands a New Security Model
19mPrivate AI: Building Trust Through Verifiable Computation
28mAI, Oracles, & Blockchains: Solving the Unstructured Data Problem
