The Toxic Alliance

One-liner: Two people who already run these tactics recognize each other, stop competing, and start trading cover, favors, and targets.

Also known as / related terms: Colluders (the “toxic triangle” model’s term for self-interested followers of a destructive leader, borrowed here for peers); mutual cover; back-scratching; collusion (the organizational-corruption literature’s term for corrupt acts that need more than one person). “The Toxic Alliance” itself is this guide’s own descriptive name, see the Label note.

What it is

Not a crowd, and not a recruit. Two operators who have recognized each other.

Most of this guide’s social entries describe one operator and the people around them: the need-driven recruit in Flying Monkeys, the drawn-in crowd in Coalition-Building / Mobbing, the protected favorite in The Golden Child. This entry is the case those leave out. Two people (sometimes three) who each already run these tactics on their own notice the same thing in each other, and stop treating each other as competition.

What forms is a working arrangement between peers. They do not aim their tactics at each other. They cover for each other. Favors move in both directions. And when their interests line up, they point the same tactics at the same people, from two directions at once.

The nearest academic anchor is the “colluder” in Padilla, Hogan and Kaiser’s 2007 toxic triangle, the model that says destructive leadership needs a leader, susceptible followers, and a permissive environment together. The model splits followers into two kinds. Conformers go along passively, out of unmet needs, low self-regard, or fear.

Colluders take part actively and for their own gain. Padilla’s own 2008 restatement with Paul Mulvey describes colluders as ambitious people pursuing selfish goals who often conspire with destructive leaders, and a Hogan Assessments summary of the same model describes them as ambitious, selfish, and sharing the leader’s worldview.

This guide’s Flying Monkeys entry is about the conformer and deliberately set the colluder aside. This entry is that colluder, with one honest adjustment: the toxic triangle is a model of leaders and followers, and this entry moves the idea sideways, to two peers who each qualify. The model itself does not describe that sideways version.

The research on organizational corruption comes closer to the peer case. Blake Ashforth and Vikas Anand’s 2003 review of how corruption becomes normal inside organizations describes groups that protect their own through unwritten rules, quoting a police-culture study whose rules include never giving up a colleague and never getting involved in a colleague’s affairs, and a shared habit of excusing each other that, repeated inside the group, turns private excuses into what the group treats as plain fact.

Jonathan Pinto, Carrie Leana and Frits Pil’s 2008 paper separates corruption that benefits the organization from corruption that benefits individuals, and treats individuals colluding for their own benefit, at the organization’s expense, as a more extreme case of the second kind.

They also note that much of the research on self-serving misconduct does not look at collusion at all. So the honest summary is that the peer-level, self-serving alliance is recognized in the research but only thinly studied.

What the research does and does not say about “finding each other”

The two strongest claims people make about this pattern are that such people find each other, and that they read each other’s moves accurately. Both are partly supported, and both have limits worth knowing.

Finding each other. Among couples, there is real evidence of similarity. A 2024 study by Gareth Richards and colleagues, with a meta-analysis attached, found small-to-medium positive correlations within romantic couples for each of the Dark Triad traits (Machiavellianism, narcissism, and psychopathy), and evidence that people seek out partners who are similar from the start rather than growing alike over time. Among peers, the picture is different. A 2016 study of 181 military cadets by Ville-Juhani Ilmarinen, Jan-Erik Lönnqvist and Sampo Paunonen found that platoon-mates with similar levels of manipulativeness or egotism were more likely to like each other mutually, but only at the low, honest end of those traits. Manipulative cadets did not prefer each other as friends. And Peter Jonason and David Schmitt’s 2012 studies found the Dark Triad traits associated with choosing friends for strategic reasons. Read together, this fits the pattern better than the folk version does: the two people in a toxic alliance do not need to like each other. The arrangement is transactional, which is exactly why it looks like mutual back-scratching rather than friendship.

Reading each other. The best-documented piece is close monitoring. In a 2014 experiment with 150 students playing a group money game, Tamas Bereczkei and Andrea Czibor found that players high in Machiavellianism adjusted their contributions to how the others in their group were behaving, especially to how many generous players were present, and finished with more money. The authors read this as Machiavellians watching their partners closely and adjusting. Note what it was aimed at, though: exploiting the cooperative players, not recognizing fellow operators. A small 2024 study by Masui, Yoshizumi and Nakajima, 170 Japanese participants judging composite photographs of faces, found that men higher in dark traits were somewhat better at inferring Machiavellianism from composite male faces, while women showed the opposite pattern. That is an interesting signal from one culture and one unusual method, not a finding to lean on. And recognizing a manipulator is not a skill only manipulators have: in a 2008 study by Minna Lyons and Sue Aitken, ordinary participants rated high-Machiavellian strangers as less likely to cooperate. The most reasonable reading, and it is this guide’s reading rather than a tested result, is that people who run these tactics recognize them in others the way anyone recognizes a move they have made many times themselves.

What it looks like (workplace)

Before: Two team leads in neighboring groups, each already known for taking credit cleanly and managing upward hard. For their first year they are rivals, and each occasionally undercuts the other in reviews.

The incident: In a cross-team review, one of them flags a real risk in the other’s project, then, after a short conversation in the hallway, withdraws it: on reflection, not an issue. A few weeks later the favor comes back. The other one backs their headcount request over a peer’s stronger case.

The month after: A colleague who raised the same risk finds her concern being described in both teams as “friction,” in almost the same words. Neither lead criticizes the other in any meeting again. When one of them is questioned about a missed deadline, the other is the first to vouch for them. Nothing any single person did was dramatic. The pattern is the pair.

How it actually unfolds

The roles here are two operators (occasionally three), whoever their shared interests run into, and the organization whose rules decide how long the arrangement lasts.

1. Recognition. Each sees the other run a move they already run themselves: credit taken without fingerprints, a colleague sidelined through the right channel at the right moment. The monitoring research above suggests this kind of close reading is real. What no research shows is a reliable “radar,” so this stage is better understood as familiarity than as a special sense.
2. Truce. The first exchange is usually something withheld: a criticism not made, a mistake not surfaced. Non-aggression costs nothing and can always be denied, which makes it the natural opening move.
3. Trade. Favors start moving both ways: support in a meeting, a vote, a piece of information, a reference, silence about a problem. Ashforth and Anand describe how quid-pro-quo arrangements drift, step by step, from overlooking small things to abetting larger ones. Economist Johann Graf Lambsdorff adds the part that holds the pair together: once two people have traded favors that could not survive being seen, each holds damaging information about the other, and both are, in his word, “locked-in.”
4. Shared targeting. When their interests line up, over the same promotion, the same critic, or the same inconvenient reviewer, both aim at the same person. Two voices arriving from two directions read to everyone else as two independent judgments, which is where this pattern often becomes the core that a wider Coalition-Building / Mobbing campaign forms around, carried by Triangulation and sometimes a Smear Campaign.
5. Mutual cover and a shared story. Each vouches for the other, and their explanations start to converge: the critic is “difficult,” the two of them are the ones who “actually get things done.” This is the group excusing habit Ashforth and Anand describe, now running between two people.
6. Fracture. The arrangement lasts only while it pays both sides. See the next section.

The alliance is conditional

This is the honest counterweight to the idea of an unbreakable pair, and it is better supported than the idea of mutual recognition. In a 2018 trust-game experiment with 80 students, Linda Szíjjártó, Ferenc Kocsor and Tamás Bereczkei found that players higher in Machiavellianism returned less of what a partner had entrusted to them, a small effect in a small sample.

David Sloan Wilson, David Near and Ralph Miller’s 1996 review in Psychological Bulletin describes Machiavellianism as broadly similar to the “defect” strategy in game theory, successful in some situations and not in others. And Lambsdorff’s analysis of corrupt deals makes the structural point: because each partner can expose the other, the same lock-in that holds the pair together is also what ends it.

His paper notes that insiders are often among a prosecutor’s most important sources, and describes an internal financial network at a large company that worked for a while, until one member was pushed out and exposed it.

In workplace terms: each member of a toxic alliance is also a potential future target of the other. The pair tends to hold while their interests point the same way, and tends to break when a restructure puts them in the same race, when one of them becomes a liability, or when the favors stop balancing.

The Fickle Operator / Fair-Weather Ally is what one partner looks like from the other’s side on the day it breaks.

Why they do it

• It pays better than rivalry. Two operators competing spend their effort on each other. Allied, they redirect it outward. This is the colluder’s motive in the toxic triangle model: personal gain through association.
• It is cheap and deniable. Nothing is written down. A withheld criticism, a well-timed endorsement, a “no concerns from my side” are all ordinary workplace acts on their own.
• It lowers each person’s risk. Each becomes the other’s witness, character reference, and early warning system.
• The environment allows it. A 2021 study by Annebel De Hoogh, Deanne Den Hartog and Frank Belschak found that Machiavellian leaders’ abusive behavior showed up where the rule climate was weak or the climate was instrumental, self-interest first, and did not show up where rules were strong or that instrumental climate was low. Ashforth and Anand similarly note that protective subcultures form more easily where, among other conditions, people are held accountable for goals but not for how they reach them. A toxic alliance is partly a symptom of what an organization rewards and what it fails to check.

What this is not

Two colleagues who back each other are not an alliance against anyone.

Coalitions are an ordinary, well-studied part of working life. Murad Mithani and Jonathan O’Brien’s 2021 review of six decades of research describes an organizational coalition simply as people who keep their differences but work together toward a goal that benefits them all. Colleagues who support each other, vouch for each other’s good work, mentor, or are simply friends are doing something normal and usually healthy.

This entry describes something narrower, and it needs all three of these to be true at once:

• Both people already ran these tactics before they found each other. The alliance joins two existing patterns. It does not create them.
• What they trade is cover for conduct that would not survive being seen, not help with legitimate work.
• Someone outside the pair pays for it.

If any of the three is missing, this is probably not what you are looking at. Two people who agree with each other in a meeting may simply both be right. Being outnumbered is not evidence of coordination.

It is also worth saying plainly that this guide describes patterns, not people. It is meant to help you recognize a dynamic you may be living inside, not to label colleagues. Once the idea of a hidden alliance takes hold, every ordinary agreement can start to look like a conspiracy, and that way of seeing has its own cost; Hypervigilance Outlives the Job is about exactly that cost.

How to protect yourself

An alliance runs on favors nobody wrote down. Make the written record the thing it cannot trade.

• Keep specific, dated facts. A pair’s main advantage is two voices telling one story. A contemporaneous record does not care how many people are telling the other version; see Document Contemporaneously.
• Don’t treat their agreement as two independent opinions. If you are the one weighing a complaint, ask whether each person formed their view separately, from their own direct experience, or whether one account is repeating the other’s phrasing.
• Don’t try to split them by carrying stories between them. That is the Divide and Conquer tactic itself. Running it back at them makes you part of the pattern and hands them a shared grievance.
• Build lines that do not run through either of them. Peers in other teams and a relationship above your manager’s level give you a channel the pair does not control; see Cultivate Skip-Level Allies.
• Remember the arrangement is conditional. You often do not have to outlast two people, only the conditions that keep them allied. That is a reason for patience and a clean record, not a reason to stay somewhere that is harming you.
• If you lead the structure, hold people accountable for how results are reached, not only for the results; avoid review setups where two people are each other’s only check; and record decisions where they are made. Lambsdorff’s own conclusion about corruption applies here: it is better fought by making corrupt relationships unstable than by relying on individual morals or penalties alone.
• If you recognize yourself here: notice whether what you trade with a colleague is silence about things that would not survive being seen, and who outside the two of you is paying for it. The alliance tends to end the way the research above describes, with one partner exposing the other, so the exit is usually cheaper the earlier you take it.

Cross-links: Flying Monkeys, the need-driven recruit this entry is not, and the entry that set the colluder aside; Coalition-Building / Mobbing, the wider group campaign a pair often sits at the core of; Divide and Conquer, the mirror image, where one operator keeps everyone else from comparing notes while the allies compare theirs; The Fickle Operator / Fair-Weather Ally, what one partner looks like on the day the arrangement stops paying; The Dark Triad, the trait cluster most of the research above measures; The Golden Child, the vertical version, favor traded for loyalty down a reporting line; The Meeting After the Meeting, the side channel where the trades usually happen; Document Contemporaneously and Cultivate Skip-Level Allies, the two defenses that do not depend on how many people agree with each other; Hypervigilance Outlives the Job, for when looking for alliances starts to cost more than it protects.

Label note: Descriptive tier, with real academic anchors. “The Toxic Alliance” is a name this guide coined for this entry; it is not an established clinical or research term, and no single study defines the peer-level pattern described here. The pieces it is built from are documented separately: the self-interested colluder in the toxic triangle model (a leader-follower concept, extended here to peers), collusion and protective group norms in organizational-corruption research, and experimental work on Machiavellian monitoring and reciprocity. The evidence on dark-trait similarity is mixed and depends on the relationship: positive and small-to-medium for romantic couples, absent at the dishonest end for peer friendships in the one field study found. The idea that such people reliably recognize each other at work is this guide’s reasonable reading, not a tested finding, and the “What this is not” section above is part of the entry, not a disclaimer.