4Gentic Research · Trust Protocol
Negotiated Autonomy is a protocol where AI agents and humans collaboratively define, execute, and evolve the boundary of autonomous action.
The agent proposes. The human adjusts. The system learns. The boundary lives.
The idea
Children gain autonomy in small increments. Teenagers renegotiate it constantly. Adults operate with broad discretion — inside webs of obligation to family, teams, and institutions. Across the whole arc, autonomy expands, shrinks, shifts, and is sometimes renegotiated from scratch.
Yet the AI industry still talks about agent autonomy as a toggle. The human is in the loop, or out. The system is a tool, or it's autonomous. We trust it, or we don't.
The real question is never "is this AI autonomous?" It is: what kind of autonomy, in this domain, under what constraints, with what consequences if it gets things wrong?
Negotiated Autonomy is our answer to that question — as an architecture, not an aspiration.
The problem
Agentic AI is moving into domains where consequences are irreversible — clinical workflows, financial operations, housing decisions, education, mental health support. Every deployment we've studied ends up choosing between:
The agent acts without sufficient judgment. In high-stakes work, a wrong autonomous action can't be corrected after the fact — the damage is done before anyone notices.
The agent asks about everything. The human develops escalation fatigue and starts rubber-stamping. You get the overhead of oversight with none of the substance.
An exhaustive decision tree that is brittle on day one, wrong the moment it meets a scenario its author didn't anticipate, and a full-time job to maintain.
The common flaw: all three treat the boundary between agent autonomy and human oversight as something you configure once. In reality that boundary is dynamic — it should differ by context, by account, by relationship, and evolve with demonstrated competence and changing conditions.
The stance
The industry's autonomy frameworks — maturity levels, governance toolkits, five-level classifications — share a structural limitation: they are top-down. The organisation defines the levels, sets the gates, and decides when to promote the agent. The agent is graded, but it never participates in defining its own operating boundary.
That's like managing an employee by job description alone, without ever asking what they think they can handle. It discards the most valuable signal available: the agent's own assessment of where its judgment is reliable — and where it is not.
In Negotiated Autonomy, the agent is an active proposer with a voice in scoping its own boundary — not a passive subject being graded until it is promoted.
The protocol
The agent reads a human-authored Goal File: the desired outcome in plain, declarative language. What good looks like — not instructions, not procedures.
The agent decomposes the goal into decision scenarios where trade-offs exist, roleplays both sides, and honestly assesses whether it had the context and judgment to make the right call.
The agent proposes a Trust Contract: "Here's what I'm confident handling autonomously. Here's where I'd check with you. And here's what I don't understand yet" — with its reasoning shown.
The human adjusts: widen this boundary, your reasoning was solid; tighten that one, there are considerations you can't see. A genuine negotiation, in both directions.
Every candidate action is checked against the contract — act, act-and-flag, or escalate. Corrections and confirmations accumulate as amendments. Evidence rewrites the contract, not single corrections.
When amendment patterns shift, the protocol distinguishes trust drift from goal drift, re-simulates against reality, and proposes a revised contract. Trust in unaffected areas is preserved.
Feedback dynamics
High-stakes agents run for weeks and months while the world shifts around them. A static autonomy classification — however well-calibrated at launch — degrades over time. So the protocol runs three nested feedback loops, each feeding the one above it.
Seconds → minutes
A contract check before each action decides: act, flag, or escalate — based on current contract terms.
Days → weeks
Amendment capture from corrections and confirmations evolves trust boundaries on demonstrated outcomes.
Weeks → months, or on drift
Goal reassessment triggered by amendment patterns detects environmental change and re-simulates the contract.
Inside the runtime
A structured, machine-readable map of the agent's competence — not a permissions list. Each entry carries a decision category, scope, a confidence score that moves with evidence, an action policy (act, act-and-flag, or escalate), and the provenance of the boundary. It also governs inquiry — how much the agent may investigate before deciding, and which questions are off-limits regardless of how useful the answers would be. Proposed by the agent, negotiated with the human, amended through operation, re-generated when goals shift.
Works backwards from the Goal File to find decision points where objectives collide, then has the agent roleplay both sides and self-assess. Over time it grounds scenarios in actual operational history, not hypotheticals.
Middleware in the critical path of every decision. Validates autonomous actions against the contract, upgrades grey-zone actions, and checks whether an escalation has already been answered before interrupting a human. Runs locally — no network round-trip in the decision loop.
A local, persistent record of every escalation and its resolution. Prevents asking the same question twice; high-reuse resolutions become candidates for promotion into the contract. Decision data never leaves the deployment environment.
Distinguishes trust drift (miscalibrated competence boundary) from goal drift (the world changed). Incremental amendments are normal learning; concentrated spikes trigger a scoped re-simulation. Trust in unaffected areas is preserved.
Routes a small, configurable sample of pre-approved autonomous actions — 2% by default — back to the human as verification probes. An independent confirm/disconfirm stream that keeps drift signal alive in even the most trusted categories.
See it in action
An interactive walkthrough of the reference runtime over a simulated 100-day horizon:
The technical paper
Intellectual honesty
Negotiated Autonomy today is a specified protocol and a reference runtime — not a completed industry standard. The fast loop is exercised in our prototype; the medium and slow loops are specified and wired into the runtime, but their empirical validation at production scale is ongoing work. The walkthrough uses modelled data, not measured production data.
We are publishing the framework precisely to invite criticism, replication, and pressure-testing. Criticism is more valuable to us than praise.
Provenance
Negotiated Autonomy was originated by Jeh Kazimi at 4Gentic — the Trust Contract, Scenario Simulator, Escalation Interceptor, Resolution Store, and three-loop architecture were first set out in a 4Gentic whitepaper in April 2026, and he later added the Verification Sampler during prototype development. A formal technical paper is in preparation: Juan Aguilar Guisado contributed substantially to its formalization, literature positioning, technical writing, and refinement, and Andres Jimenez Ramirez served as an internal guide and critic throughout development. The broader 4Gentic team contributed by testing the practical implications across the company's ventures.