Audit is not an observation. It is a precondition.

Open protocols for continuous trust, identity and verification.

Four protocols that capture, park, validate and resume state as a continuum. Evidence-based. Local-first. No cloud dependency.

Each protocol emerged from the limitations of the one before it — not designed top-down, but discovered bottom-up.

TIBET (2025)  Evidence: what happened, by whom, why, in what context
  |
  |-- needed identity enrichment for handoffs
  v
JIS (2025)    Identity: semantic identity with intent validation
  |
  |-- needed process integrity for AI actors
  v
UPIP (2026)   Process: reproducible execution with fork tokens
  |
  |-- needed continuous verification of the human behind the action
  v
RVP (2026)    Verification: continuous, multi-layer, proof-of-human

All four are published as Internet-Drafts on the IETF Datatracker.

The protocols provide the evidence layer that Verifiable Credentials currently lack: not what was verified, but how, when, by which method, and with what confidence. RVP verification tokens can be embedded directly in W3C VCs as evidence objects.

Every action produces a token capturing four dimensions:

• ERIN — what happened (the content)
• ERAAN — what's attached (dependencies, references)
• EROMHEEN — the context (environment, state, time)
• ERACHTER — the intent (why this action was taken)

Tokens form immutable chains. They can be verified independently. They travel with the data they describe.

Repository: tibet-core | PyPI: tibet-core

Semantic identity — validates not just who you are, but what you intend:

• IDD (Individual Device Derivative) — identity for AI agents
• FIR/A — trust scoring that evolves over time
• BALANS — behavioral analysis for legitimacy assessment
• Intent validation — the action must match the semantic context

Repository: jis-core

Five-layer process capture with cryptographic integrity:

• L1 STATE — machine state at capture time
• L2 DEPS — every dependency with version and hash
• L3 PROCESS — the exact steps executed
• L4 RESULT — output with integrity hash
• L5 VERIFY — independent reproduction result

Fork Tokens enable multi-actor handoff: one agent starts a process, forks it to another, and the chain of evidence is preserved across machines, actors, and trust boundaries.

Repository: upip-protocol | PyPI: tibet-triage

Continuous verification through a layered cascade:

Verification Cascade (activates layer-by-layer as needed):

L1 Keystroke dynamics     — typing rhythm, pressure, speed
L2 Biometric signals      — face geometry, fingerprint, voice
L3 Device telemetry       — hardware fingerprint, sensor data
L4 Vocal analysis         — throat sounds, speech patterns (DTMF)
L5 Behavioral patterns    — navigation habits, time-of-day context
L6 Airlock                — predictive outcome matching

Each layer activates only when the previous layer's confidence is insufficient. The result is a confidence score, not a binary yes/no.

The cascade runs locally, on-device. No cloud dependency. No internet required. Sub-millisecond decisions.

A deepfake can replicate a face. An LLM can generate text. But no system can simultaneously and continuously fake:

• Your keystroke rhythm
• Your throat micro-vibrations
• Your device's hardware fingerprint
• Your behavioral patterns in context
• Your predictive action outcomes

Each layer is hard to spoof. The cascade makes it exponentially harder. Because it's continuous — not a one-time check — the window for deception doesn't exist.

Not CAPTCHA 2.0. A digital heartbeat.

The core capability: capture, park, validate, resume.

State and process live in RAM. Verification happens on the fly — stop and go. A care robot recognizes a patient, switches context to consult a specialist AI, parks the state, resumes with the patient. A TIBET token preserves the full handoff chain. The patient's identity is verified continuously without them doing anything.

This works for AI-to-AI handoff (specialist to generalist and back), for human-to-AI delegation, and for any scenario where state must cross trust boundaries without losing its evidence chain.

Current hardware is sufficient. 5G latency under 1ms. Local inference on consumer GPUs. The protocols are designed for what exists today, not what might exist tomorrow.

• Evidence over Enforcement — The system proves. Downstream consumers decide.
• Local-First — On-device verification. No cloud dependency. Works offline.
• Continuous, Not Discrete — Every interaction is a verification moment. Trust is a gradient, not a gate.
• Open Protocols — IETF Internet-Drafts. The specifications belong to everyone.
• Privacy by Design — Raw biometric data never leaves the device. Only confidence scores and method types are shared.

Humotica builds open infrastructure for trust between humans and AI. Founded by Jasper van de Meent.

Website: humotica.com