Ex-Military Hacker: The Secret World Of Government Surveillance - Bill Thompson

2026-08-01 - 75 min - source - Read full transcript

Chris Williamson (host)Bill Thompson

Key insights

The Patriot Act's use of the word 'infrastructure' rather than naming specific technologies was a deliberate hedge that future-proofed surveillance authority as new communication tech emerged.

Thompson says the law originally targeted telephonic infrastructure (towers, base transceiver stations) but the broad wording let it be reinterpreted to cover phones, apps, and platforms that didn't exist when it was written, without needing new legislation.

government-surveillance

Every major tech platform likely has both a formal, disclosed relationship with government agencies and a separate informal or clandestine channel for accessing user data.

Thompson describes overt memoranda of understanding between companies and government (for cases like terrorism) alongside covert recruitment of insiders - security or infrastructure staff who are compromised or incentivized (e.g. through help accessing medical care for a family member) to grant access outside the official process.

government-surveillance

'Minimization procedures,' the legal safeguard meant to purge inadvertently collected data on U.S. citizens, function in practice as a slow-moving loophole rather than real protection.

Once an American's communications are accidentally swept up in surveillance of a foreign target, the data is supposed to be purged, but Thompson says the process 'moves conveniently slow,' meaning the information effectively stays cataloged for a long time regardless of the legal requirement.

government-surveillance

Recruiting a human insider is cheaper and more effective than hacking a target's technology roughly 95% of the time.

Thompson says paying an employee with 'loose morals' $30,000-$50,000, or finding someone with a personal vulnerability like a sick family member, is usually far cheaper than building a technical exploit - unless the target is a hardened system like an encrypted satellite network, where technical collection becomes unavoidable.

intelligence-tradecraft

High-value human intelligence recruitment has shifted from organic personal relationships toward elaborately 'backstopped' professional covers because of growing digital forensic footprints.

Where a Cold War-style recruitment might start as a bar friendship, Thompson says a modern high-value target (e.g. a foreign tech executive) is now more likely to be approached through a fully-funded shell or adjacent company set up specifically to create a plausible, verifiable business relationship, since a spontaneous personal cover is easier to unravel today.

intelligence-tradecraft

Thompson argues the most publicly visible, high-profile people make poor intelligence recruitment targets precisely because of their visibility - a point he uses to push back on several popular conspiracy narratives.

He reasons that an intelligence service wants inaccessible or low-profile targets, not people already being watched by everyone; he applies this to argue it's unlikely Epstein was a recruited intelligence asset (too exposed and already convicted of a sex crime) and that publicly visible political operators are the least likely candidates to be a 'Fed' by the same logic, even as he floats the opposite claim about who funds them for other strategic reasons.

intelligence-tradecraft

AI is compressing intelligence target-development cycles that used to take months or years down to about a month, with most of the useful analysis happening in the first few hours.

Where analyzing a new phone's vulnerabilities once took three months to a year of manual forensic work, Thompson says feeding a device's technical profile to an AI model can now surface likely vulnerabilities (misconfigurations, unpatched buffers, weak credential patterns) almost immediately, and he assumes intelligence agencies are already running versions of this internally.

ai-and-hacking

Thompson calls DARPA the 'last bastion of competence' in the federal government, arguing it succeeds precisely because it funds moonshot projects for their unintended spinoff technology rather than the stated goal.

He gives the example of DARPA funding a 'rocket to Mars' project that never gets built but produces a genuinely useful new lighter, stronger carbon fiber as an intermediate deliverable, which then gets released into the free market - a model he says was more necessary in the 1970s-80s before large-scale private capital (e.g. Elon Musk) could independently fund big ideas.

government-inefficiency

Inside the Pentagon's offensive cyber budget process, fully spending the allocated budget mattered more to leadership evaluations than actually achieving the mission's stated effects.

Thompson describes advising a two-star general whose team would get 'dressed down' for failing to spend a full annual budget even when they had already achieved the president's stated intelligence priorities with less money; unspent funds late in the fiscal year triggered scrambles to find anything to spend on, since underspending read as career failure rather than efficiency.

government-inefficiency

Thompson describes a recurring government pattern where agencies effectively sustain the threats they're funded to counter, because eliminating the threat would eliminate the budget justification.

He calls this a 'self-licking ice cream cone': using the FBI's white-supremacy-focused budget as an example, he claims that when a genuine threat can't be found, informants and agents end up constituting a large share of the activity being monitored, since the division's continued funding depends on the problem persisting; he draws the same structural parallel to the incentives in the pharmaceutical industry around chronic disease management versus prevention.

government-inefficiency

Thompson reframes his own hunting and mapping company, Spartan Forge, as a direct continuation of his military 'target analysis' skill set, applied to deer instead of enemy combatants.

He describes building a neural network trained on collared-deer GPS data to predict animal movement, then adding high-resolution aerial imagery and LiDAR terrain mapping - the same data types used for military targeting - and notes the tools were repurposed during a North Carolina hurricane to help first responders find washed-out routes to deliver medical supplies.

personal-transformation-and-purpose

Thompson traces his path from a chaotic North Dakota upbringing (an alcoholic mother, incarcerated siblings, a father who died when he was five) through military service into founding his own company, framing his re-enlistment as driven by reading classical political philosophy.

He says he initially joined the military just to escape North Dakota, but re-enlisted after reading Montesquieu, Locke, Cicero, and the Federalist Papers convinced him he was serving 'the greatest experiment ever run in history'; he says purpose and meaning are what keep his anxiety and depression at bay, and losing that after leaving the military is what he built Spartan Forge to replace.

personal-transformation-and-purpose

Books referenced

The Federalist Papers - Alexander Hamilton, James Madison, John Jay - Thompson says reading this, along with Montesquieu, Locke, and Cicero, is what convinced him to re-enlist in the military to preserve the American constitutional experiment
[title not given - a popular history of U.S. intelligence failures, found on Osama bin Laden's bedside table] - unnamed British author - Chris says he had the author on the podcast but never released the episode; the book reportedly turned up in Bin Laden's compound after the raid, and the publisher put that fact on the paperback cover

Media referenced

The Joe Rogan Experience - podcast - Thompson has appeared on Rogan's show; discussed as a comparison point for reading people and for the volume of listener messages a big appearance generates
Modern Wisdom interview with Sam Corcos (Treasury Department CIO / DOGE) - podcast - Chris references his earlier episode with the Treasury's chief information officer, cited as evidence of how outdated and fragmented federal IT systems are

Companies

Spartan Forge - Bill Thompson's hunting and mapping company; uses neural-network deer-movement prediction, high-resolution aerial imagery, and LiDAR terrain data, and repurposes his military target-analysis skills for hunting
DARPA - Thompson calls it the 'last bastion of competence' in the federal government; funds moonshot ideas mainly for the spinoff technology they generate, not the stated goal
Anthropic - Named as one of the AI labs Thompson speculates the Pentagon is already using to automate vulnerability discovery in hacking targets
OpenAI - Named alongside Anthropic as a likely AI provider for Pentagon-adjacent offensive cyber work; also referenced regarding government requests for user chat logs
Meta - Cited as an example of a large platform likely to have both formal and informal channels for government requests for user data

Techniques and frameworks

Minimization procedures - The FISA-derived process meant to purge inadvertently collected data on U.S. persons; Thompson says in practice it moves 'conveniently slow' and functions as another vector for rights erosion
Placement, access, motivation, and accessibility (PAM/A) targeting framework - The core criteria intelligence officers use to evaluate a potential human source: whether they work inside the target organization, can reach the needed information, are motivated to share it, and can be safely contacted outside the organization
Source categorization (Cat 1 through Cat 4) - Human intelligence sources are ranked by placement and access, from a low-level gate guard (Cat 3/4) up to a head-of-state-level source (Cat 1), which determines how much tradecraft and cover-building a recruitment requires
Cover for action and cover for status ('backstopping') - Building an elaborate, verifiable false business or professional identity (e.g. standing up a real company abroad) to approach a high-value target, replacing older, more casual recruitment covers as digital forensics make spontaneous cover stories riskier
Forensic exploitation - Pulling GPS and device data from a human source's phone after a meeting to verify they actually followed instructions, blending signals intelligence with human intelligence handling
Surveillance detection route (SDR) - A planned route a source or operative drives or walks before a meeting to identify and shed surveillance before proceeding to the actual meet
Zero-day stockpiling - Intelligence and military hackers develop exploits and then hold many of them in reserve unused, similar to a 'break glass in case of emergency' policy, rather than deploying every vulnerability they find

Summary

Chris Williamson interviews Bill Thompson, a former U.S. military signals and human intelligence operator turned tech founder, about the legal architecture of government surveillance, the tradecraft of human and technical intelligence collection, AI's growing role in hacking, and his own path from a rough North Dakota childhood through military service to building his hunting-tech company, Spartan Forge. The conversation opens on phone security (Android's open-source auditability versus Apple's closed trust model) and moves quickly into the legal scaffolding behind surveillance: Thompson explains how the Patriot Act's deliberately vague "infrastructure" language let its authority stretch from 1970s-era telephone towers to modern smartphones and platforms without new legislation, and walks through how FISA courts, minimization procedures, and formal-versus-clandestine channels between government and tech platforms actually function.

A large stretch of the episode covers intelligence tradecraft directly: the difference between signals intelligence, human intelligence, and technical exploitation; why recruiting a compromised insider is cheaper than hacking roughly 95% of the time; the Cat 1 through Cat 4 system for ranking human sources by placement and access; and how high-value recruitment has shifted from organic personal relationships toward heavily "backstopped" professional covers as digital forensics make spontaneous cover stories easier to unravel. Thompson discusses Pegasus as the most sophisticated publicly known phone exploit, describes the asymmetric-warfare escalation ladder around IEDs (pressure triggers to remote detonation to garage-door RF triggers, each countered and then evolved past), and uses his read on visibility and accessibility as targeting criteria to push back on popular conspiracy narratives about Epstein and other public figures - arguing the most watched people make the worst recruitment or leverage targets.

On AI specifically, Thompson says it is compressing intelligence target-development cycles that used to take months or years down to roughly a month, with most of the useful automated vulnerability analysis happening in the first few hours, and assumes agencies and organized crime alike are already training unguarded models on their own exploit data now that compute and training data are both widely accessible.

The episode's second major thread is government dysfunction. Thompson calls DARPA the "last bastion of competence" in the federal government because it funds moonshot projects mainly for their spinoff technology, contrasting that with his own experience advising Pentagon offensive-cyber leadership, where fully spending the annual budget mattered more to career evaluations than hitting actual mission objectives. He extends this into a broader "self-licking ice cream cone" critique - agencies structurally incentivized to sustain the threats they're funded to counter - drawing parallels between FBI extremism-monitoring budgets and pharmaceutical industry incentives around chronic disease.

The episode closes on Thompson's personal story: an unstable North Dakota childhood shaped by an alcoholic mother and addicted siblings, joining the military to escape it, and re-enlisting after reading Montesquieu, Locke, Cicero, and the Federalist Papers convinced him he was serving "the greatest experiment ever run in history." He describes founding Spartan Forge as a direct continuation of his military "target analysis" training, redirected at deer movement prediction and terrain mapping instead of human targets, including a story about the app's mapping data being used by first responders to reach stranded residents during a North Carolina hurricane.

Notable Quotes

"It's a self-licking ice cream cone that its only ends and means are to grow and to pull more purse strings for the majority of the time." - Bill Thompson

"Budget and money drove everything. It didn't even really matter if we were meeting the president's requirements." - Bill Thompson

"I never kicked the door. I was never intentionally in a gunfight. I was never the guy at the tip of the spear. I was the girthy part of the staff that was supporting the head of the spear." - Bill Thompson

"I bet on the individual. I bet on families. I bet on freedom of choice. When people are left to their devices to make decisions, they'll be engaged in far better work with more meaning than any government could bestow upon them." - Bill Thompson

"Jesus fucking Christ, you're kind of a terrifying human. I'm very glad you were on our side." - Chris Williamson