Skip to content
Regulayer™Human Control for AI
Book a live demo

Human authority in the loop

It’s not human in the loop. It’s human authority in the loop.

Human authority in the loop means a named person’s current authority, checked at the moment the AI acts, and a sealed record an outsider can check. The work keeps flowing. A human stays in control.

How it works

Set the authority once. It is checked at every action.

  1. 1A person sets the authority once. What the AI may do, how much, and for how long.
  2. 2Regulayer™ checks it before each action goes through. In milliseconds. No one has to approve each step.
  3. 3Outside the authority, it is sent back for correction. If correction fails, the action holds and does not run.
  4. 4Withdraw the authority, and the same request stops. Every decision, allowed or stopped, is sealed as it happens.

The difference

Human review decides. Regulayer™ makes that decision enforceable when the AI acts.

 Human in the loopHuman authority in the loop
The questionDid a person approve this step?Does a named person’s current authority cover this exact action?
When it is checkedApproval systems ask a person again, or rely on a standing grant.Regulayer™ continuously checks whether the named human authority is still in force when an action would take effect.
The personApproves step after step, and becomes the bottleneck.Sets the authority once. No one has to approve each step.
A change of mindA grant given earlier can still be relied on after the person’s authority has changed.Withdraw the authority, and the same request stops.
The AI is talked into itA person has to spot it in time.The model may agree. The action still requires current human authority.
The recordKept inside the system that acted.A sealed record an outsider can check, without Regulayer™.

The evidence

A person on every step is the bottleneck.

The published record on step-by-step human approval, every figure linked to its source.

KPMG AI Quarterly Pulse · Q1 202663% of large US firms now require a person to validate AI agent outputs, up from 22% a year earlier.237 leaders at firms with $1 billion or more in revenue
Clinical decision support · systematic reviewDrug safety alerts were overridden in 49 to 96 per cent of cases.van der Sijs et al., JAMIA 2006, 17 studies
Controlled experiment · PLOS ONE 2024A human-in-the-loop design raised uptake of algorithmic advice and lowered decision accuracy.Sele and Chugunova, N=292
Gartner · prediction · June 2025Over 40 per cent of agentic AI projects cancelled by the end of 2027, citing costs, unclear value or inadequate risk controls.Gartner press release

Why it matters now

Regulators, platforms and insurers now require a human in control of AI.

US FDA · first warning letter on AI · 2 April 2026“Any output or recommendations from an AI agent must be reviewed and cleared by an authorized human representative.”The record in the Consequence Library
EU AI Act · Article 14Effective human oversight, and the ability to intervene or stop the system.Regulated industries and GxP
FDA and EMA · Good AI Practice · January 2026Data provenance, processing steps and analytical decisions documented in a traceable and verifiable manner.Principle 6, in the FDA’s own text
Texas Department of Insurance · Bulletin B-0003-26 · June 2026Where AI is used to make a consequential decision, a person reviews and agrees with the decision before action is taken.The rules on insurers’ AI
Bermuda Monetary Authority · draft guidance · August 2026For higher-impact agent actions, controls should not rely solely on prompts or model instructions.The rules on insurers’ AI
California SB 947 · Chapter 859 · 2026No discipline or termination decided by an automated system alone. A human must corroborate the output.The law, section by section

A term from the actuaries

Actuary-in-the-loop.

Institute and Faculty of Actuaries · AI Manifesto · 24 September 2026A qualified person checks the AI’s output, the way actuaries check models.Alex Waite, IFoA President-elect and the manifesto’s author: actuaries “have been checking models for nearly 180 years.” A check needs evidence of what the AI was authorised to do and what it did. Regulayer™ supplies that evidence. The AI Manifesto

See it

The work keeps flowing. A human stays in control.

Regulayer™ supplies the evidence. Your auditor, regulator or court makes the determination. Patent pending.