Join War on the Rocks and gain access to content trusted by policymakers, military leaders, and strategic thinkers worldwide.
The U.S. military spends billions trying to predict what its adversaries will do — and almost nothing testing whether its own plans make sense. In part, this asymmetry reflects the decline of red teaming, the practice of systematically challenging plans to expose biases, blind spots, and weak assumptions before they become operational liabilities. As U.S. adversaries become more sophisticated and the Department of Defense integrates AI into planning, restoring systematic self-critique isn’t optional — it’s urgent.
Restoring red teaming will require updating how it was practiced in the past. There are four possible ways to make this happen: train all planners — not just specialists — in assumption-testing, focus on building red team capabilities rather than red teams, strengthen mechanisms for including red teaming in ongoing planning cycles, and use forecasting methods and platforms to trigger assumption reviews.
Targeting Failure Pathways
Red team methodology primarily targets three failure pathways in defense and military planning. Biases like optimism bias or mirror-imaging creep in when planners unconsciously assume the adversary will behave as they would, or that their preferred outcomes are more likely than they are. Blind spots emerge from organizational or institutional silos. Planners may simply not see how logistics, allies, civilians, or political dynamics could shape the battlefield in ways they haven’t anticipated. The most dangerous failure pathway is unevaluated assumptions — uncertain claims upon which entire plans depend, but which rest on weak arguments, weak evidence, or both.
Integrating AI into planning amplifies these risks, as it can encode biases in its training data, create blind spots when algorithms function as black boxes, and generate confident outputs based on unvalidated assumptions. As these systems become embedded in planning, systematic assumption-checking becomes more, not less, critical. For a time, the U.S. military invested in mitigating these risks, but then it stopped.
I served on the U.S. Central Command Red Team in 2012–2013, when many combatant commands maintained in-house red teams. Our mission goal was simple but uncomfortable: expose flawed logic before it hardened into the command’s war plans. Red teaming was never just “thinking like the enemy.” Instead, it was an internal audit of thought that mapped assumptions, surfaced blind spots, and forced planners to confront what might go wrong.
Red Teaming’s Rise and Fall
After the U.S. invasion of Iraq in 2003, commanders took stock of planning failures, such as overestimating Iraqi security force capacity, failing to anticipate how de-Baathification could fuel insurgency, and underestimating the rise of sectarian violence. To avoid repeating those mistakes, they began embedding structured contrarian analysis into planning through red teaming. Building on older “devil’s advocate” and wargaming traditions, the Army spearheaded formalizing the practice, founding the University of Foreign Military and Cultural Studies in 2004 to teach planners how to recognize cognitive bias and challenge assumptions. Within a decade, many combatant commands had adopted similar methods, making red teaming a routine feature of the joint planning process.
Doctrine eventually caught up. Joint Doctrine Note 1‑16 and later Joint Publication 5‑0, both in the main text and an appendix, encouraged commanders to use red teams for independent review. Yet, even as the concept gained legitimacy on paper, institutional support waned. The Army’s red team school closed in 2021 without a designated successor, leaving red team education fragmented across service schools and online modules. Many commands retained “red cells” focused on adversary emulation, but discontinued standing red teams that probed their own reasoning.
Over time, the phrase “red team” became more branding than practice. Cyber units still use it to test network defenses, and private industry borrows it for security audits. But the broader application of challenging operational and strategic assumptions has withered. A 2024 Data & Society study observed that organizations often “enroll red teaming for optics,” as a ritual of accountability rather than a mechanism of learning.
The result is predictable. With fewer institutional challenges, planners focus intensely on an adversary’s actions but neglect their own assumptions. This produces a pattern that analysts keep identifying: plans built on shaky assumptions that haven’t been stress-tested before operations begin.
Listen to this episode of The Insider featuring author Alexandra Gerber by becoming a War on the Rocks member here
Why Red Teaming Met Resistance
Red teaming declined for understandable reasons: time constraints, resource limits, and resistance to being challenged. Organizational behavior theory helps us understand how organizations can be resistant to change. Planners under tight deadlines tend gravitate to what seems most plausible, most digestible, and the least contentious. A red team was in the room to slow that reflex, to ask the questions others could or would not, and to provide deliberate pushback. During my time at Central Command, I watched commanders and planners struggle with uncomfortable analysis even when they recognized its validity.
Two key examples from my time on the red team illustrate the challenge. In 2013, command plans for Afghanistan and Syria rested on two core assumptions: Afghan security forces would be able to secure the government and prevent the Taliban from retaking control following the end of the International Security Assistance Force mission and President Bashar al-Assad’s regime would soon collapse due to the Syrian civil war that began in 2011. Our team challenged both assumptions, arguing that Afghan forces might not be capable and that Assad’s hold on power could prove more durable than expected. Planners acknowledged these alternative scenarios but struggled to incorporate them into formal plans. There was no clear process for developing formal contingency plans without an explicit order directing that work, no incentive to complicate an already complex planning effort, and no requirement to present the alternatives to the decision-makers.
Planners, and the commanders they work for, often operate under compressed timelines and ask legitimate questions such as: Can we afford to slow down planning to question our assumptions? Won’t constant challenge undermine cooperation in the planning team? If red teams identify problems but the system can’t accommodate alternative plans, what’s the point of the exercise?
These objections deserve answers. First, the concern that red teaming creates delay rests on a flawed premise. Checking your work as you go isn’t waste: it’s prudent planning and the only way to make best guesses as to whether plans can withstand implementation. Consider the costs of discovering during execution that partner forces can’t sustain operations, logistics assumptions don’t hold, or political constraints were miscalculated. Those kinds of failures lead to months or years of costly adjustment. Red teams exist to shift those discoveries left on the timeline — from the battlefield to the briefing room — where course corrections are possible without casualties or crippling expense.
Second, the concern about undermining the collaborative planning environment misreads legitimate challenge for sabotage. Red teams don’t exist to second-guess planners or commanders. Rather, they exist to stress-test logic before it hardens into operational plans — to ensure confidence is based on examined assumptions rather than untested optimism. That’s not undermining leadership. It’s supporting it.
Finally, the complaint that “red teams identify problems, but nothing changes” points to a system problem rather than a red team problem. When a red team’s findings reach commanders only as optional briefings rather than mandatory decision points, and when planners face no requirement to document why they accept or reject alternative planning scenarios, warnings become noise. The U.S. Intelligence Community has faced similar challenges with handling alternative analysis. The 2015 revision of Intelligence Community Directive 203 addressed this problem by requiring intelligence products to identify core assumptions, present credible alternatives, and document why those alternatives were set aside. That model could readily be applied to planning: require red team findings to be presented to commanders and require planners to document their rationale for setting aside red team findings.
Another common objection to red teaming is something along the lines of “we already do extensive wargaming — isn’t that enough?” It’s not. Wargaming tests plans against an adversary’s courses of action, revealing vulnerabilities in execution. Red teaming challenges the logical foundations and assumptions underlying those plans, revealing flaws in conception. A plan can perform well in a wargame while resting on faulty assumptions about partner force capacity, logistics, or political constraints. Both forms of challenge are necessary, but neither is a replacement for the other.
Dissenting views rarely die from bad evidence — they die from lack of process. Without institutional mechanisms to channel alternative analysis into decision cycles, uncomfortable insights get swept aside. If speed is consistently prioritized over ensuring validity, the ability to anticipate failure atrophies. But that can be fixed.
Why Red Team 2.0 is Needed
To avoid the next failure of imagination, several practical reforms could re-anchor red teaming in the planning process to better reflect the current realities of joint operational planning and warfighting.
Make Red Teaming Everybody’s Business
Existing requirements in joint doctrine could be strengthened so that joint planners are trained in how to include assumption-testing in their planning process. Planners, like everyone else, are already pressed for time, but they should have at least some understanding of the main failure pathways — biases, blind spots, and unevaluated assumptions — especially as AI-assisted planning tools that can encode hidden biases or rest on unverified data become commonplace. Senior planners should have even more robust training so they can lead planning efforts that address these issues. Initial red team training could be as little as a two hour module in basic planner training, whereas more advanced planners attending senior Army or joint planner training would need greater exposure.
Study and Resource What Works
While every command needs red team capabilities, not every command needs a permanent red team. The right red team model depends on command-specific planning characteristics. Commands doing continuous, high stakes contingency planning for complex operational environments — typically geographic combatant commands like European Command or Indo-Pacific Command — would benefit from a permanent, in-house red team capability to review plans as they evolve. By contrast, commands with more episodic planning cycles or narrower technical focus, such as Africa Command or Space Command, may only require external reviews during major exercises or plan updates. Service component commands might rely on planners with some red team training rather than building a dedicated red team. Matching requirements to solutions requires empirical assessment — each combatant command, or the Department of Defense as a whole, should study which approaches work best under what conditions to guide appropriate investments.
Link Red Teaming to Ongoing Planning Cycles
Planning isn’t a single event — it’s a continuous process. Operational planning teams often revisit and update plans as conditions change. Every review should include a mandatory assumption audit, with core assumptions tracked and updated each cycle. Building periodic assumption checks into the planning rhythm would make challenge the norm rather than the exception. Red teaming complements wargaming: While wargames test plans against adversary actions, assumption audits stress-test the logical foundations of the plans themselves.
Integrate Forecasting Methodologies
Forecasting can serve as a trigger for red teaming by signaling when assumptions should be reconsidered, such as by identifying shifts in partner capacity, changes in logistics delivery timelines, or regime stability. While forecasting may not deliver the precision that planners are looking for, it can be a useful tool for identifying when assumptions no longer align with reality. Tools such as the RAND Forecasting Initiative could be woven into planning cycles to provide early warning when key assumptions begin to diverge from observed trends.
Red teaming didn’t fail — it became inconvenient. But plans built on convenient assumptions don’t survive contact with reality. As adversaries grow more capable and planning integrates AI, the cost of unexamined biases and blind spots will only increase. The U.S. military built red teaming capabilities for good reason, and those reasons are more valid than ever. Red teams were never meant to make planning comfortable. Irritation was the point. Better that friction arise in the planning room than on the battlefield.
Alexandra Gerber is a researcher at RAND and a graduate of the U.S. Army’s Red Team Short Course.
**Please note, as a matter of house style, War on the Rocks will not use a different name for the U.S. Department of Defense until and unless the name is changed by statute by the U.S. Congress.
Image: Sgt. Jose Escamilla via DVIDS
