Independence
Built in, and open to check.
Our work is only useful to a regulator if it is independent, and independence is only credible if it can be checked. This page sets out how ours works: what we will not take, what we will not sell, and what we will not do. It is a commitment we expect to be held to.
We are commercial, and independent of our clients
Greymark Research is a commercial company. We sell intelligence, research, and training, and we make no apology for being paid to do serious work. The payment is for the work, and it earns no say over what the work finds. We hold no ownership stake, equity, or investment relationship with the organisations we report on or sell to, and no client can direct a finding. Our work answers to the evidence, because the evidence is the only thing we are selling.
We are independent of the AI labs we examine
The platforms we study are built by frontier AI labs. We take no funding, grants, compute credits, equity, or partnership from any lab whose platforms fall within our subject matter. This is a structural rule, written down and applied, because a regulator cannot rely on evidence produced by an organisation funded by the sector it is regulating. What stays open is the transfer of intelligence itself. We will brief regulators, publish our methods and findings, and let our work be cited freely.
Child-safeguarding intelligence is never sold
Where our work surfaces specific, actionable evidence of risk to children, that intelligence is not monetised. It goes first to the relevant regulator, authority, or safeguarding body, free of charge. Commercial products may discuss child-safety-relevant trends, regulatory exposure, anonymised patterns, and generalised risk pathways, but they do not sell identifiable safeguarding evidence, child-risk leads, or operational details that should be escalated as a protection concern.
The ecosystem cannot be its own customer
Our commercial intelligence maps the open-source AI, crypto-rail, anonymous-inference, and grey-market ecosystems. An operator we map cannot buy the product that maps it. For this purpose, an operator means an organisation whose own services, infrastructure, platforms, models, payment rails, repositories, inference endpoints, or companion or chatbot products are directly mapped, scored, or assessed in the relevant commercial product. This is a written operational rule, applied before any sale and the same way every time.
We name on the evidence, and we never expose private individuals
Naming public actors where the evidence warrants it is legitimate, legally protected, and a standard part of open-source intelligence work. We do it neutrally, and only where the evidence supports it. We do not, and will not, reveal personal information that a person has kept out of public view. We do not doxx, in any project, for any client, at any price.
Our published research is open
Our published research is free to read, and we host it where it lives, on the journals, repositories, and preprint servers that hold the record. Open publication is part of how independence is checked. Anyone can read the work and judge it for themselves.
If you believe we have fallen short of any commitment on this page, tell us
This statement was last updated in June 2026.