ThirdKey — Zero Trust for AI
001 — Open research

Specifications, not slides.

ThirdKey publishes the trust primitives we ship as open specifications and peer-reviewable preprints. Every cryptographic boundary, every policy phase, every guarantee — written down so security teams can audit it before it runs in production.

002 / Preprints

Public research.

Open-access on Zenodo. Specifications, threat-research preprints, and empirical evaluations — each one a self-contained artifact a security engineer can read, implement, and audit.

Preprint May 6, 2026

VectorSmuggle: Steganographic Exfiltration in Embedding Stores and a Cryptographic Provenance Defense

An empirical study of how attackers with write access to RAG ingestion pipelines can hide payloads inside embeddings — noise injection, rotation, scaling, offset, and fragmentation that preserve retrieval behavior while defeating statistical detectors. Disjoint-Givens rotation reaches ⌊d/2⌋·b bits per vector (1,920 bytes at d=3072). Introduces VectorPin — Ed25519 signatures binding each embedding to source and model, breaking verification on any post-embedding tamper.

VectorPin Ed25519 RAG security Steganography Embeddings
Specification May 1, 2026

ToolClad: Declarative Tool Interface Contracts for Agentic Runtimes

A specification for declarative tool interface contracts in agentic runtimes. A single .clad.toml manifest defines typed parameters, validation, invocation, output parsing, and Cedar policy — across CLI, sessions, and governed browser execution. Replaces freeform shell generation with allow-list-validated tool invocation.

.clad.toml Tool contracts Cedar Allow-list MCP
Preprint April 26, 2026

Typestate-Enforced Agent Loops: Making Policy Gates Unskippable at Compile Time

A compile-time approach to enforcing policy gates in AI agent loops via typestate encoding. Evaluated across nine hosted LLM providers — 263 forbidden tool-call attempts refused without execution at 30–95µs per check. Addresses the time-of-check-to-time-of-use problem with affine ownership semantics.

Typestate Affine types Cedar Rust OATS
Specification April 17, 2026

Open Agent Trust Stack (OATS): A System Specification for Zero-Trust AI Agent Execution

An open specification for zero-trust execution of AI agents. Defines declarative tool contracts with allow-list enforcement, compile-time verification of the Observe-Reason-Gate-Act loop, and structural separation of the policy gate from language-model influence across five architectural layers.

Zero-trust ORGA loop Tool contracts MCP Cryptographic identity
003 / Applied writing

Recent posts.

Working notes from research.thirdkey.ai — what we’re building, what we got wrong, what changed.

May 9, 2026

VectorSmuggle: What Embedding Stores Trust, and Why That’s a Problem

RAG ingestion pipelines treat any vector that retrieves well as legitimate — a trust model attackers exploit to hide payloads inside embeddings. VectorPin closes the gap with Ed25519 signatures binding each vector to source content and producing model.

April 13, 2026

Your Agent’s Guardrails Are Rotting

Prompt-based safety degrades under operational load — a kind of context rot that parallels human goal neglect. The fix is structural: policy evaluation and typed contracts, not prose rules.

April 3, 2026

Stop Letting Your Agent Write Shell Commands

ToolClad replaces freeform shell generation with a declarative manifest of typed parameters and templates — safer agent execution through allow-list validation, not deny-list filtering.

March 2, 2026

ORGA: A Typestate-Enforced Agent Runtime That Makes Policy a Phase, Not a Feature

A Rust agent runtime that enforces policy evaluation as a mandatory compile-time phase — combining typestate patterns, durable journaling, and cryptographic auditing so agents can’t bypass authorization.

View the full archive on research.thirdkey.ai ↗
A guarantee that can’t be written down isn’t one.
— ThirdKey, research principle