Pentera Turned a 12-Attack Beta Into a $100M ARR Business by Making Security Prove Itself

The most expensive sentence in cybersecurity may be, “We think it works.” A company buys firewalls, endpoint agents, identity controls and a scanner that returns a festive spreadsheet of red cells. The scanner can tell the team what might be weak. It cannot always tell them whether an intruder could stitch those weaknesses together, cross the network and reach the account that opens everything. Pentera built its business in that uncomfortable gap between having security and proving security.

The Burlington, Massachusetts-based company, founded in Israel as Pcysys, sells an automated security-validation platform. With authorization and guardrails, it behaves like a controlled burglar. It discovers an opening, tries an attack technique, collects evidence, moves laterally when permitted and maps the path toward a valuable target. Then it gives the defenders a ranked repair list. The point is not to create one more alarm. The point is to show which alarm connects to an actual route through the building.

01 / The first failureA scanner found too much and proved too little

Traditional penetration tests are useful, but they are appointments. A consultant arrives, tests a defined slice of the environment and delivers a report. The network changes the following week. New accounts appear. Cloud permissions drift. A firewall rule is adjusted because somebody needs a Friday deployment to ship. The handsome PDF remains frozen in time.

Vulnerability scanners solve the frequency problem but create a prioritization problem. They can produce thousands of possible flaws, often sorted by a generic severity score. A security team still has to decide whether a medium-rated configuration and an exposed credential combine into a route to domain administrator. Pentera’s thesis was that the attack itself could do the sorting. If a weakness participates in a working attack chain, it moves up the queue. If it cannot be reached or exploited under the tested conditions, it can wait.

That difference places Pentera between several established categories. It overlaps with breach-and-attack simulation vendors such as AttackIQ, Cymulate, Picus and SafeBreach; automated pentesting platforms such as Horizon3.ai’s NodeZero and RidgeBot; attack-path tools such as XM Cyber; vulnerability-management suites; and human pentesting services. Pentera’s pitch is more active than a model and broader than a one-off engagement: run controlled attacks across internal networks, external assets, cloud identity and hybrid infrastructure, then carry validated findings into remediation.

02 / The wedgeLet the buyer watch the product break in

Arik Liberzon, who had led a red team in the Israeli military, founded the company in 2015. The first beta arrived in 2017 with roughly a dozen common Windows attacks. It was not a miniature version of today’s platform. It was a narrow claim with a difficult constraint: automate the work without knocking over the customer’s production environment.

Prospects doubted that fully automated penetration testing was possible. Pentera did not answer with a longer presentation. It offered qualified buyers a free, single-day proof of concept in their own environment. The software was agentless, so the team could get to evidence quickly. Early tests reduced a haystack of possible weaknesses to the small portion that could participate in a breach. Customers watched the system take advantage of misconfigurations and weak credentials in minutes or hours.

“Belief alone doesn’t move analysts, execution does.”Amitai Ratzon, CEO

That was the mind-changing moment. The buyer no longer had to accept a startup’s theory about a new category. The buyer had an attack path inside a familiar network. Blackstone became an early customer, then an investor in the 2019 Series A. The relationship gave Pentera more than capital: it supplied a reference inside the trust circles of large American enterprises. The reusable lesson is sharp. When buyers doubt that a new type of product can work, design the sales motion so the product creates a fact they can repeat to their boss.

03 / The machineFour surfaces, one awkward question

Pentera Core tests internal networks, credentials and lateral movement. Surface looks from the outside at internet-facing assets. Cloud validates cloud identity, infrastructure and the seams between cloud and on-premises systems. Resolve, created from Pentera’s 2025 acquisition of DevOcean, pushes downstream: it organizes validated findings into remediation workflows. The company also sells advisory work and SECTOR11 adversarial services, expanded through the acquisition of EVA Information Security, for AI red teaming, applications and tests that benefit from human judgment.

The product is sold as enterprise software through direct sales, partners and marketplaces. It fits offensive-security teams, red teams, vulnerability managers and CISOs who need evidence they can show to a board. Public customer references include Blackstone, DTCC, Wyndham Hotels & Resorts, Casey’s and Telefonica. In one DTCC account, the platform was described as an amplifier for a lean team, particularly for repeatable ransomware testing and executive-ready reporting.

At that price point, the math has to include more than software seats. A buyer is comparing it with external pentest fees, red-team labor, the time spent validating scanner output, and the opportunity cost of fixing the wrong things. Pentera claims reductions in third-party testing costs and remediation time, but the most credible business case is local: run a proof of value, compare the attack paths with known findings, measure staff hours and decide whether the team can act on the output.

The model scales because the same attack knowledge can be sold repeatedly as subscription software. Pentera Labs updates techniques; integrations send evidence into tools such as ServiceNow, Jira and major security platforms; channel partners help with deployment and operation. By January 2026, Pentera said more than 1,200 enterprises in over 60 countries used the platform. The company also reported passing $100 million in annual recurring revenue in 2025, with roughly 450 employees across 20 countries.

04 / The turnFinding the fire was not enough

The company’s product expansion reveals a useful tension. Validating a breach path creates urgency, but it can still end as another ticket in a crowded queue. DevOcean gave Pentera the machinery for Resolve, which adds business context, ownership and workflow to the repair process. EVA added expert-led testing for AI systems and applications. Together, the deals moved Pentera from “show me what breaks” toward “help me close it,” while acknowledging that software does not own every testing scenario.

In 2026, Pentera pushed further into AI-guided offensive work, web application testing, an MCP server for AI SecOps workflows and a partnership with Recorded Future. That last connection is especially tidy: current threat intelligence enters the platform, the system tests whether the threat works in this environment, and the team receives evidence instead of a generic warning. Pentera also became available through AWS Marketplace, shortening procurement for customers that already buy through AWS.

05 / The copyable playbookSell the proof, then earn the platform

Founders can borrow five parts of the Pentera playbook. Start with a repeated expert workflow, not the whole profession. Keep the first library narrow enough to control. Remove setup friction from the proof of value. Demonstrate inside the customer’s real context. Finally, make the result easy for the champion to retell: a working attack path is a far better internal story than a feature matrix.

There is a second lesson in the sequence of products. Pentera did not begin by promising one platform for every attack surface and remediation task. It earned the right to expand. Core created evidence. Surface and Cloud widened coverage. Resolve followed the result into the customer’s workflow. Services filled the spaces where human ingenuity still matters. The platform is the accumulation of proven adjacent jobs, not a noun pasted onto the homepage.

Automation also introduces its own trust problem. A product allowed to attack production needs permissions, scoping, containment, audit trails and cleanup that buyers can examine. A false negative can breed confidence; an unsafe action can cause damage. Pentera’s early insistence on safety was therefore not just engineering hygiene. It was the condition that made its category sellable.

The company’s most interesting achievement is not the unicorn valuation or the funding, though it has raised about $250 million. It is the behavioral change behind the revenue. Pentera persuaded enterprises to treat penetration testing less like an annual ceremony and more like a recurring check. The controlled burglar keeps returning, because the building keeps changing.