Friday afternoon. A director opens Tuesday's board pack. In it: an independent valuation from a name-brand firm, a market benchmark cited in the strategy paper, a consultant's options paper that points to a clear preferred direction, and an advisory committee minute endorsing it. None of these documents discloses, on the page where the director encounters them, who paid for the work, who picked the panel, or what material stake the producer has in the outcome the document quietly supports. The director reads them as inputs into a decision. They are not inputs. They are products. Products of an architecture.
There is one question I bring to any document of consequence that lands on a leader's desk, before any other question. Not whether the author is competent. Not whether the firm is reputable. Not whether the methodology survives scrutiny. The question is the simpler one. Whose salary was the writer drawing while writing this?
I should say at the outset that the question is not popular. Most readers of documents like the ones in that board pack are not, in the end, looking for the truth. They are looking for usable leverage. A document is read once for conclusions that can be quoted in the next meeting, once for caveats that can be safely ignored, and not at all for the question of who paid for it to exist. The salary question is unwelcome precisely because the answer might complicate the document's usefulness as cover. That is captured human nature operating at the level where it does the most damage. Not corruption. Convenience. The architecture I am about to describe survives because the demand for cover meets the supply of capture exactly where the calendar needs it to.
So I am writing for the smaller audience. The leader who reads a document and wants to know whether it was written to help them think or written to help someone else win. That audience is small. It is also where the work gets done.
Where the architecture is clearest
The cleanest version of this architecture right now sits inside mainstream journalism, where the published reporting has been most thorough. The Tarbell Center for AI Journalism, a San Francisco non-profit founded in 2022, pays the full salaries of early-career journalists placed inside Bloomberg, TIME, The Guardian, NBC News, The Verge, MIT Technology Review and several others, for nine-month placements. Stipends run from sixty thousand to one hundred and ten thousand US dollars. The host newsroom pays nothing. The funder pays everything. The byline carries the publication's name. The reader sees the publication. The reader does not see the funder.
The Tarbell Center's principal donor is Coefficient Giving, which until recently was named Open Philanthropy, the philanthropic vehicle of Facebook co-founder Dustin Moskovitz and his wife Cari Tuna. Its other principal donors are tied through various vehicles to the Effective Altruism movement, which has spent the last decade advocating specific regulatory and safety regimes around frontier AI development. The journalists Tarbell pays the salaries of are journalists covering artificial intelligence in TIME and Bloomberg and The Guardian. A Washington Examiner review of the 2025 cohort's published output found the topical signature you would expect: AI safety, AI regulation, AI-induced harms in mental health, in misinformation, in employment, in the energy grid. Not corruption. Topic selection.
The Pulitzer Center's AI Accountability Network operates a comparable fellowship programme funded principally by George Soros's Open Society Foundations alongside the MacArthur Foundation, Omidyar Network, Luminate and others. The Omidyar Network's Reporters in Residence programme places stipends with freelance journalists covering technology and society. The architecture varies. The pattern repeats. Across these programmes, the count of journalists working inside one or another version of this arrangement since 2022 is approaching one hundred. None of this is disclosed to the reader at publication in the way the byline is.
The lead cases happen to cluster around foundations associated with one political colouration because that is where the public reporting has clustered. The same diagnosis would apply, line for line, to a foundation associated with the other colouration funding the salaries of journalists at publications associated with the other colouration. The architecture is what it is regardless of which direction the funder leans.
What it sells in this particular instance is worth naming. The funders' stated position is that advanced AI poses an extinction-level risk and that regulation, governance, and safety constraints around frontier development are urgent. My position differs and I have made the longer argument elsewhere. The short version is that the human response to advanced technology is adaptation, not legislation against it, and what the foundations are actually selling under the heading of "AI safety" is the same fear-dressed-as-ethics architecture humanity has always built when it has lost faith in its own capacity to evolve. Techno-morality, as I called it in Evolve or Be Remembered, is just institutional fear with better UX. Ethics laundering. Compliance theatre. The Tarbell architecture is the distribution mechanism for the product. The reader who picks up the Guardian piece on AI safety is not just being informed. They are being sold a worldview, by a journalist whose monthly salary was paid by the people selling it, inside a publication whose byline is loaning the worldview the credibility of independence.
That is the press dimension. I want to deal with it here as a leadership matter, because the architecture I have just described is the same architecture every operator in this country is sitting inside, reading from, and producing for.
The same architecture, in your own house
Some of what follows will sting. You have done some of it. You have allowed some of it. You have acquiesced to some of it. That is not blame. The architecture is older and larger than any single leader. But where you have been executive in producing it, facilitating it, or accepting it, the accountability is yours.
Some of this list has crossed your desk. Some of it more than once. You will know which.
You have read sponsored research published with the sponsor's name absent from the cover. You have read think-tank reports written inside funding arrangements whose direction of travel was set at the donor level rather than the analyst's desk. You have signed off on board papers whose annexes were produced by advisers with material stakes in the recommendation under discussion. You have read consultant findings, cited as independent, produced by firms whose next engagement depended on the answer. You have sat on advisory boards drawn from grateful suppliers and presented as independent advice. You have published thought leadership under your own byline that was substantially written by an external party with a position. You have attended conferences whose panels and topics were curated by the sponsoring company. You have circulated industry analysis at trade events that was, in the part of the document that mattered, a supplier-funded white paper.
Some of this you have produced. Some of this you have commissioned. Some of this is sitting inside your organisation right now, ready to be circulated next week under one or another institutional letterhead.
The long lever
A salary is a long lever. A foundation that pays a journalist's salary for a year does not need to direct a single sentence to shape what gets covered, what gets foregrounded, what gets called important. The architecture does the work. The funder picks who gets hired. The training programme frames what counts as a serious question. The summit decides which sources are credible. The career path runs through the network. The human mind, in any line of work, attends to the questions the surrounding environment frames as the important ones.
The same lever runs through the leader's own document supply. Who picked the consulting firm sitting on the panel? Who selected the analysts who wrote the benchmark study now sitting in the deck? Who funded the institute whose framing is now in the strategy paper? Who sits on the advisory board that signed off on the recommendation? Whose conference did the idea come from, and who paid for the room? At every step, the lever operates upstream of any single sentence. By the time the document reaches your hands, the framing has been done. The questions worth asking have been pre-selected. The conclusions defensible inside the architecture have already been narrowed.
The strict-firewall assurance, when leaders give one, answers a question that is not the question. Whether any single piece of analysis was directed by the funder is rarely the issue. The architecture directs the work without ever needing to direct a sentence. That is what architectures do.
The structural choice
The diagnosis applies symmetrically inward. The salary question is now yours to ask, of every document your organisation produces, sponsors, commissions, or quietly relies on. The question is not whether you, the leader, are honest. Honesty is the lowest bar in this conversation and clears no part of the problem. The question is whether the architecture under which the document was produced is one you would object to as a reader of someone else's.
If it is, the choice is structural, not motivational. There is no call to authenticity here. Either the paymaster gets named at the top of the document, in the same typeface and at the same weight as the author's name, or the document does not go out. Either the advisory board includes voices with no material stake in the recommendation, or the disclosure of that stake sits on the face of the recommendation. Either the consultant who will do the next engagement is not the one writing the case for the next engagement, or the conflict is named in the cover note. Either the executive who is going to be quoted in the press release wrote what is in the press release, or someone notes that they did not. Either editorial independence inside the institution is a structural discipline, with reporting lines and incentives that protect it, or it is a line in the values section of the website.
The choice does not need to be performed. It only needs to be made.
The time problem
The objection, if it has not been formed already, is the obvious one. How is any of this supposed to be found out when there is a business to run, payroll to make, and a finite number of hours in the week? The honest answer is that the audit alone is not the answer. I do that audit myself on what crosses my own desk, and have become rigorous about it. AI verification tools and disciplined speed-reading let me reach a factual position on what I rely on before I rely on it. It still costs something even done this way. I have not always done it. I do it now because the cost of not doing it has become higher than the cost of doing it. The personal audit, however rigorous, does not scale beyond one desk. The scalable answer is structural. Building the institution so the daily audit is not needed across hundreds of desks is the work. It looks like two-line attestations at the head of any commissioned document, naming who paid for it and what material stake the producer has. It looks like standing rules on who can sit on an advisory board and on what terms. It looks like reporting lines that make disclosure cheaper inside the organisation than hiding. It looks like disclosure by default rather than disclosure on application. The discipline runs as a system. The leader does not police it daily because the system makes the captured document harder to produce than the clean one.
There is also a personal answer, and it bears saying because the question is honest enough to deserve one. I am writing this knowing what the couple of hours I spent researching, verifying, and writing it cost me in revenue-generation time. I am making a bet that the work being read here, by the smaller audience I named at the outset, returns more than the same hours spent chasing a fee would have returned. The bet may not pay off on any single article. It pays off, if it pays off, on the cumulative standing built for what I will and will not say. The reader making the corresponding bet, that the audit of their own document architecture is worth more than the alternative use of that attention, is making the same kind of bet I am making. Either it pays off or it does not. Pretending the question does not apply because the calendar is full is itself a choice. It is not a clever one.
What the people around you are reading
The captured architecture in the press is consequential because the public extends the press a trust calibrated for an earlier architecture. The captured architecture inside your own institution is consequential for the same reason. The people who depend on your judgement, your staff, your board, your investors, your customers, your peers, are extending you the same calibration. They are reading you the way the public used to read mainstream journalism. They assume the byline is the writer. They assume the recommendation is the analyst's. They assume the independent review is independent.
The architecture survives because most readers prefer it to survive. That is the part of this conversation nobody wants on the page. Once it is on the page, the survival of the architecture inside your own house becomes a choice you are making, not a condition you are inheriting. The leader who has now seen it can see it. The leader who can see it has the choice. That is the only question on the page. Everything else is decoration.
Paul Lange advises owners, executives, and boards on the decisions that define commercial outcomes and organisational character, and on what a working board actually contributes. He has spent close to four decades across finance, technology, hospitality, professional services, and operating roles, in Europe, Asia, the Middle East, and Australia, on both sides of the table, with private equity and venture capital one part of it, and has taken five of his own companies through to exit. He is the creator of the Total QX™ and TILE Theory™ frameworks, and the author of The 20% Leader, Mis(très)s Entrepreneur Manifesto, Evolve or Be Remembered, and The Inheritance Manifesto. He runs his advisory practice, Manolutions, from the Gold Coast, Queensland. He writes Conviction because leadership without accountability is just theatre.