This is the shortest entry in Law 31's list and the only one that can put the reader in prison.
This is a classic con-artist technique: You attract your victims to some criminal scheme, creating a bond of blood and guilt between you. They participate in your deception, commit a crime (or think they do—see the story of Sam Geezil in Law 3), and are easily manipulated.1
The case:
Serge Stavisky, the great French con artist of the 1920s, so entangled the government in his scams and swindles that the state did not dare to prosecute him, and "chose" to leave him alone.1
And then the instruction, stated flatly:
It is often wise to implicate in your deceptions the very person who can do you the most harm if you fail. Their involvement can be subtle—even a hint of their involvement will narrow their options and buy their silence.1
Recorded as written, without softening and without argument.
The technique is: identify whoever has the power to destroy you, and make them a criminal. That is the content. Greene names it as a con-artist method, gives a con artist as the case, and recommends it in his own voice with the word wise.
This page carries it because the vault documents what sources actually say. It is worth being exact about what is on the page, because the surrounding entries — a diplomat's memos, an art dealer's naps — set a register this does not belong to.
Note also the escalation in the final sentence. "Even a hint of their involvement" extends the method from actual complicity to manufactured appearance of complicity, which is a different and worse act: framing someone who did nothing, in order to buy their silence.
The other six methods narrow someone's options by arranging what they can see, want, or reach. This one narrows their options by creating a permanent fact about them.
That has consequences none of the others have.
It does not expire. Kissinger's menu ends with the meeting. Vollard's buyer leaves the shop. A crime does not stop having happened.
It is symmetrical. Every other technique leaves the operator in a stronger position than the target. This one leaves you both holding the same weapon. The government could not prosecute Stavisky; Stavisky could not stop being a man the government wanted gone.
It creates an interest in your death rather than your defeat. A target who has been out-manoeuvred can retreat. A target who is only safe while you are silent has a permanent incentive that ordinary opposition does not produce.
Greene's own framing concedes the first of these and none of the others: the bond is "of blood and guilt," which is a phrase about permanence.
The chapter stops at "the state did not dare to prosecute him." That is where the technique looks like it works.
What Greene does not say — and the omission is the entire evaluation — is what happened next.
The Stavisky affair broke in 1934. The scale of the entanglement became public, Stavisky was found dead in a chalet in Chamonix in January of that year (officially suicide, widely disbelieved), and the scandal brought down a French government and triggered the 6 February 1934 riots, one of the most serious crises of the Third Republic.
[SPECULATIVE] — none of this is in Greene's text; it is standard history and is flagged as coming from outside the source, not from it.
The technique's failure mode is now visible and it follows directly from the structure. A mutual hostage arrangement holds only while the exposure is contained. The moment it becomes public — and Stavisky's scale made that inevitable — every implicated party's incentive flips at once from protecting him to destroying him fastest, and the operator is the single point at which the whole thing can be resolved.
Greene stops the story at the top of the curve. On the fuller account, Brothers in Crime is not a technique with a risk; it is a technique whose success condition is its own eventual detonation.
A small structural detail worth flagging.
(or think they do—see the story of Sam Geezil in Law 3)1
This is one of the very few explicit internal cross-references in the entire book. Greene repeats cases across laws constantly, contradicts himself across adjacent chapters, and tells the same anecdote twice at different numbers — all without comment.
The one time he points the reader across law numbers is for a con-artist technique, and specifically to establish that the victim need not have actually committed a crime, only believed they did.
It is a fair summary of the corpus's centre of gravity that this is the cross-reference that earned an aside.
The honest workflow for this one is defensive, because the offensive version is a crime and the vault is not going to write instructions for it.
You are being brought into something and it is going well. Somebody capable and charming is including you, and the inclusion feels like trust.
Watch for one specific shape: being made a participant in something small and irregular, early, before there is any reason to. A signature you did not need to give. A payment routed through you. Being copied on something you have no business being copied on. Being the one who told someone the thing that turned out to be untrue.
None of these will look like the main event. They are not supposed to. "Even a hint of their involvement" is the operative phrase, and the technique works by making the entanglement seem incidental at the point it is created and load-bearing at the point it is used.
The test is not is this illegal, which is too high a bar and arrives too late. The test is: if this relationship ended badly tomorrow, is there anything here I would not want described? If yes, that thing is the hostage, and its size is irrelevant.
And the only defence is unpleasant and works: surface it early, voluntarily, to someone who is not them. An irregularity you have disclosed cannot be used. The disclosure costs you the relationship, which is the price, and it is much lower now than it will be.
Strongest evidence. A named operator, a dated decade, and a real and well-documented episode in French political history. The mechanism as described is accurate to the case as far as Greene takes it.
Tension — the case is stopped at the point where the technique looks successful. Stavisky died in 1934 and the scandal brought down a government. Documented above; [SPECULATIVE] and sourced from outside the text.
Tension — the third risky entry, unmarked, and the risk is categorically different. The other two risky methods can fail. This one can succeed and then destroy you. See The Six Methods of Controlling Options.
Tension — the final sentence escalates from complicity to its appearance. "Even a hint of their involvement" is framing an innocent party, offered as a refinement.
🚩 [POPULAR SOURCE] · 🚩 SINGLE SOURCE · 🚩 SECONDARY WITHOUT PRIMARY — no source for Stavisky, no named officials, no date beyond "the 1920s", and no mention of the 1934 collapse. The omission is not neutral: the chapter's evaluation of the technique depends entirely on where the story is cut.
Open questions. Is there a version of this that is not criminal — mutual reputational exposure, shared responsibility for an unpopular decision — and does it have the same detonation property? Ordinary organisational life is full of arrangements where two people are jointly committed to something they would each rather not have to explain, and the structure looks identical minus the felony.
Within Greene's own corpus this sits oddly against Law 2 and Law 42.
Law 2 — Never Put Too Much Trust in Friends argues that entanglement with people who know you is the standing danger. This entry recommends manufacturing exactly that entanglement, deliberately, with the most dangerous available person. The book's warning and its instruction are the same arrangement viewed from opposite ends, and neither chapter mentions the other.
Against Alter the Playing Field — the other method Greene aims at "those who resist at all costs" — the comparison is the sharpest thing in this law. Rockefeller neutralises opposition by owning the railways. Stavisky neutralises it by making officials criminals. Both are offered as answers to the same diagnostic question, one costs capital and the other costs your liberty, and the flat list marks no difference.
And Houdini versus Kleppini is the benign version, which is worth noting because it shows the mechanism separated from the crime: Kleppini is lured into a scheme he believes is dishonest — stealing a code, rigging a duel — and his own sense of complicity is what stops him complaining afterwards. Nobody committed a crime. The bond of guilt did the work without the guilt being warranted, which is the Sam Geezil variant Greene cross-references and the only defensible member of this family.
History — the arrangement that holds until it doesn't. Power Consolidation Through Orchestrated Betrayal traces regimes that bind subordinates through shared responsibility for acts none of them can disown, and the characteristic end state: the binding holds while the regime holds, and dissolves catastrophically and simultaneously the moment it does not.
Stavisky is that pattern at private scale, and the historical frame supplies the timing rule Greene's account lacks.
The insight neither gives alone: a mutual-hostage arrangement is not a stable equilibrium that occasionally breaks — it is a stable equilibrium with exactly one exit, and every participant takes the exit at the same moment. That is because the thing restraining each of them is the others' silence, so the first defection removes the restraint on everyone simultaneously. There is no gradual unwinding and no partial exposure, which means the operator gets no warning: the arrangement looks maximally secure right up until the instant it is worthless. Greene's evidence for the technique — the state did not dare prosecute him — is a description of the calm, not of the system.
Psychology — why the small first step works. Commitment and Consistency Escalation describes the sequence in which a trivial initial agreement makes a larger one substantially easier, because people work to keep their self-image consistent with what they have already done.
That is the machinery behind "even a hint of their involvement." The first irregularity is not chosen for its leverage — it has almost none — but because it converts the target from someone this is being done to into someone who has done something.
What the pairing produces is the defence, and it explains why the obvious defence fails. People try to defend against this by watching for the big ask, and the big ask is not where the commitment is made. The binding happens at a point too small to trigger a refusal, and by the time anything is large enough to notice, refusing it means also disowning the earlier thing — which people will not do, because the earlier thing is now part of who they are. So the only effective test is not about magnitude at all: would I mind this being described? applied to items far below the threshold where anyone normally applies it. And the only effective response is disclosure, because disclosure is the one move that severs consistency pressure rather than accumulating it.
Sharpest implication. Every other method in this chapter leaves the operator stronger than the target. This one leaves both parties holding the same weapon, permanently, and the arrangement has exactly one exit that everybody takes simultaneously — because what restrains each participant is the others' silence, so the first defection unbinds them all at once. Greene's evidence that it works (the state did not dare prosecute him) is a description of the calm before that, and he cuts the story in 1930 rather than 1934.
Generative questions.