About
Who authorized this action, and can an outsider check it?
Current human authority for consequential AI. AI acts. Humans authorize. Regulayer enforces, and proves it.
Generative and agentic AI become consequential when their outputs or actions affect people, systems, assets or decisions. They can keep interpreting and acting after the human decision that governed them has changed. Regulayer holds the authority that currently governs the work, decides whether the machine may proceed under it, and produces evidence of what happened.
Current human authority → check before action → independently verifiable record.
Independence
Your authority does not have to live in somebody else’s cloud.
Regulayer does not require a cloud service of its own. It runs outside the system it governs and independently of the model provider, so the authority that decides what the machine may do stays in your custody. Local, private and on-premises configurations are supported where the deployment allows, and the evidence can be verified by anyone holding the file, without an account and without contacting us.
One estate
One floor, and the surfaces built on it.
Every surface answers one of three questions: what authority governs now, may the machine proceed under it, and what is the evidence.
Current human authority
HumanMark records how the work was actually made. Presence records that a named person was there when it was signed.
Control before consequence
Drift Control holds a consequential answer until current authority resolves it. FailStop holds a consequential action. SDKontrol carries the same control into a vendor’s own product.
Verifiable proof
The Receipt is the proof object a person signs and sends. Witness is the queryable history behind it. The open verifier checks either, free, offline, independently.
Two ways to buy it
Use it, or license it.
Use Regulayer
Professionals who need evidence of what AI did and what a human reviewed. The desktop product runs on your own machine, and the first outcome does not require understanding the estate behind it.
License Regulayer
Enterprises, AI platforms, OEMs, agentic-system providers and regulated operators embed current-human-authority enforcement and independently verifiable evidence in their own systems, deployed on their own infrastructure.
Where it is put to work: AI agents, legal practice, robotics and autonomous systems, autonomous laboratories, industrial operations, finance, and other systems whose actions carry consequences.
The work behind the file
The architecture came first.
Then it stopped
Energy harvesting, distributed cognition, the physics of intelligence that does not need a server. The work was real and recorded. Then it stopped.
Not bolted on
It was not ready to release without control. A control designed into the operating architecture is stronger than one added afterward.
One act
So the architecture sat on the shelf until the control existed. Safety was not added. It was made part of the same act.
The architectural truth
You breathe in oxygen. You breathe out carbon dioxide. It is the same breath.
The kernel governs because it has to, not because it was added later. The proof is the byproduct of the inference, not assembled after the fact. The receipt cannot exist without the decision, and the decision cannot exist without the receipt.
Architecture and safety are not separate. They are one breath.
The founder
Louisa Johnson Bullock
Founder, Regulayer
A career spent where safety is never promised, it is architected and proven. A career in regulated industries, before AI: USP 797 sterile compounding, where the standard was written by a United States Pharmacopeia committee and she drove its adoption in the field, across hospital, pharmaceutical, and nuclear-pharmacy workflows. CRISPR commercialisation at Thermo Fisher, with global go-to-market for genomics, protein sciences, and biobanking across forty countries. The Tyvek clean-environment business at DuPont, product that itself had to be manufactured clean. In-line semiconductor manufacturing at Texas Instruments, with processes set up across Southeast Asia. Contamination control at ITW.
In a semiconductor cleanroom or a sterile-compounding suite, you do not trust that the environment is clean. You build it clean by architecture, and you produce the proof.
Regulayer applies a principle familiar from regulated manufacturing and life sciences: important controls are stronger when they are designed into the operating architecture and accompanied by evidence that the control operated.
Regulayer applies that principle to consequential AI. Prove it, don’t just trust it. Build the control in from the start.
The position
An enabling company, not a competitor to AI.
It is an enabling architecture for AI systems that require external authority, consequence control and independently verifiable evidence. Others are invited to build their own instances on the kernel. HumanMark is meant to be public infrastructure.
Filed, and reduced to practice
The intellectual property estate.
Priority running from 27 April 2025. Patent pending.