Bishop Fox Built a $154 Million Business by Breaking Into Companies Before the Bad Guys Do

Bishop Fox has a pleasantly awkward product to describe at a board meeting: it gets paid to break into the company. The Tempe, Arizona, firm attacks applications, clouds, networks, hardware, artificial-intelligence systems and, when invited, the human beings and physical doors surrounding them. Then it stops, writes down the route and helps the customer close it. This is offensive security, a corner of cybersecurity where permission is the line between a consultant and a criminal.

Founded in 2005 by Vincent Liu and Francis Brown, Bishop Fox grew up as a specialist consultancy. Both founders had practiced enterprise penetration testing at Honeywell. The firm still sells carefully scoped assessments, but its more consequential move was to turn an occasional expert project into a recurring operating system. Cosmos, its cloud-native platform, continuously discovers a customer's public assets, verifies what is actually reachable, gathers evidence and sends the serious candidates to Bishop Fox operators. Those humans test exploitability before a result reaches the client.

The product is a shorter argument

Cybersecurity departments do not lack warnings. They lack enough hours to investigate them. Conventional scanners can produce long lists of weak software versions, exposed services and suspicious configurations. A finding may be technically correct yet practically harmless. Another may look modest in isolation but become dangerous when chained with two other mistakes. Bishop Fox's proposition is that customers should buy the conclusion, not merely the list.

Cosmos builds a living inventory of domains, subdomains, cloud resources, network services and web applications. Stateless microservices handle the repetitive work at enterprise scale. Protocol checks verify that an asset exists and behaves as expected. Evidence such as fingerprints, screenshots and service metadata is collected. Then an operations team performs the expensive bit: it thinks like an attacker, validates the path and decides what deserves scarce remediation time.

“Technology alone doesn't solve exposure management.”Vincent Liu, co-founder and CEO

The platform is deliberately invisible. Customers do not deploy or maintain Cosmos; Bishop Fox operates it within a managed service and exposes results through a portal and integrations. ServiceNow and Jira connections push validated issues into the systems where security and engineering teams already work. This sounds like a small product choice. It is actually the business model: software creates leverage, experts create trust, and the customer pays for an outcome rather than adopting another dashboard that needs an owner.

Zoom was the stress test

The cleanest example arrived in 2020, when Zoom's daily meeting participants jumped from 10 million to 300 million in four months. Each new business account could create subdomains for employees and contractors. Bishop Fox's case study says Zoom suddenly had roughly 250,000 new subdomains to protect and about 500,000 targets across its broader attack surface.

What failed first was not a firewall. It was the calendar. A point-in-time assessment could describe yesterday's estate, but new infrastructure appeared too quickly for yesterday to remain useful. After conversations with Bishop Fox, Zoom concluded that it needed testing broader and more agile than a standard periodic engagement. Cosmos mapped unknown assets and tracked changes. Bishop Fox researchers sifted the potential issues, constructed proof-of-concept exploits and ranked what Zoom should fix first. The service complemented Zoom's bug-bounty program rather than replacing it, using bounty input as another signal for validation and priority.

This distinction matters. Bug bounties invite a distributed crowd to find individual problems. Continuous attack-surface testing tries to maintain a coherent view of everything exposed. A traditional pentest goes deep on a defined scope. Bishop Fox sells all three kinds of logic around one customer problem: the environment will change, attackers will notice, and internal teams need help deciding where to look.

What the bill is really measuring

Bishop Fox does not sell from a public price card. Engagements are scoped around assets, cadence, testing depth and the amount of expert involvement. Its published business-case model instead prices the alternative. The company estimates that recreating a continuous offensive-security capability internally could cost $2,788,750 a year: data feeds and workflow tools, plus an attack-surface analyst, vulnerability researchers and analysts, security engineers, penetration testers and a remediation manager.

The buyer is usually a mature organization with valuable systems, a sprawling internet presence and an internal security team already drowning in triage. Public customer references include Zoom, Equifax, Google, John Deere, Sonos, Coinbase, Republic Services and Reltio. The company says it has worked with more than a quarter of the Fortune 100 and eight of the ten largest global technology companies. These customers are not outsourcing responsibility. They are renting specialized offensive capacity and a repeatable testing machine.

A wide menu, one underlying trick

Bishop Fox's service catalog runs from application, mobile and secure-code reviews to cloud testing across AWS, Azure and Google Cloud. Network work covers internal, external and wireless systems. Product specialists reverse-engineer binaries and protocols, fuzz connected devices and physically manipulate hardware. Red teams combine technical intrusion with social engineering and, where authorized, physical access. Readiness work includes ransomware exercises and incident-response tabletops.

The breadth separates Bishop Fox from a pure attack-surface software vendor. Its modular red-team work can follow an adversary across email, identity, cloud and a building entrance. Its product-security people can move between firmware and the companion app. Competitors range from firms such as NCC Group, NetSPI, Mandiant, Coalfire and Cobalt to platforms from CyCognito, Pentera, Rapid7, Tenable and Qualys. An experienced in-house red team is the most direct substitute.

The company also gives away part of its expertise. Bishop Fox Labs maintains more than 20 open-source tools. Sliver is a cross-platform command-and-control framework for adversary emulation. CloudFox maps attack paths in AWS. JSluice extracts secrets and endpoints from JavaScript. Eyeballer applies computer vision to reconnaissance screenshots. Unredacter demonstrates why pixelation is not redaction. These projects solve working testers' problems, build credibility with technical buyers and create a recruiting signal that a glossy culture page cannot fake.

AI raises the volume, not the certainty

In February 2026, Bishop Fox introduced Cosmos AI inside its application-pentesting workflow. The engine explores applications at machine scale while human testers validate each reported finding. That positioning is deliberate. Liu has argued that AI makes capable researchers faster while also helping inexperienced people generate polished but inaccurate work. In other words, the same technology that expands testing coverage can multiply false confidence.

Two months later, the company released AIMap, an open-source system for discovering exposed AI services, fingerprinting their frameworks, scoring risk and launching authorized tests. It began in a hackathon and was assembled in a single afternoon from off-the-shelf components. The speed was the finding: reconnaissance that once demanded a specialist can now be built cheaply. AIMap covers exposed MCP servers, Ollama instances, model proxies and popular AI application frameworks, then visualizes them on a globe. It is both a useful tool and a neat advertisement for Bishop Fox's thesis.

The model explores. The operator decides. The customer fixes.

The part worth copying

The Bishop Fox playbook travels beyond security. Begin with a difficult expert service. Instrument the repeatable steps. Automate discovery, collection and routing. Keep judgment with people who can defend a decision. Sell a managed loop, not a pile of features. Publish useful tools to prove that the expertise is real. Most important, place the software behind the customer outcome instead of making adoption itself the customer's project.

It works when mistakes are expensive, environments change continuously and customers lack enough specialized talent. It works especially well when raw automation generates a queue larger than the team can handle. The model is weaker for a small company with a stable footprint, for a buyer that only needs a narrow compliance scan, or for an organization unable to repair the findings. Verification has little value if the engineering backlog is a graveyard.

Bishop Fox's culture fits the product. It is remote-first, spread across 19 countries and 38 U.S. states, and fond of slogans that sound like the wall of a hacker clubhouse. One official value is “Give a Shit.” Another is “Be Excellent to Each Other.” Under the jokes sits a serious quality-control mechanism: clients grant unusually sensitive access, and the firm must attract people capable of abusing it who can also be trusted not to.

After two decades, the company sits in an interesting middle ground. It is too software-driven to be a conventional consultancy and too human-dependent to be ordinary SaaS. That tension is not a defect. It is the entire pitch. Attackers combine automation, patience and improvisation. Bishop Fox has built a company that tries to do the same, only with a contract, a stop condition and a repair ticket at the end.