For a security team, the most dangerous vulnerabilities are the ones you don't know how to look for. Legacy scanners can check for common technical flaws, but modern applications built on GraphQL and sprawling microservices present a different problem. The real risk often sits in the business logic, the unique workflows and data flows that a generic tool can't comprehend. This is the gap Escape was founded to attack, not with more rules, but with AI agents that learn how an application works and then try to break it [TechCrunch, June 2023].
From an initial focus on API security, the Paris-based company has expanded its positioning into a full AI-powered offensive security platform. It aims to automate the penetration testing lifecycle, offering continuous, automated discovery and exploitation of vulnerabilities across modern applications [Escape Careers]. The premise is straightforward: if manual pentests and bug bounties are slow, expensive, and intermittent, a machine that can test like an attacker, around the clock, should be able to find more, faster.
The Wedge Was GraphQL
Escape's founding insight was specific. Co-founders Tristan Kalos and Antoine Carossio met at UC Berkeley and identified a lack of developer-oriented tools for building secure APIs, particularly as GraphQL adoption grew [Startup Savant, October 2024]. Traditional security tooling often treated GraphQL as an odd-shaped REST API, missing the nuance of its queries and mutations. Escape's initial product dynamically generated realistic requests for these modern API formats, using an AI-powered algorithm to simulate attacks and propose fixes [TechCrunch, June 2023].
This focus on business-logic testing, rather than just scanning for known technical vulnerabilities like SQL injection, became the wedge. It addressed a pain point for development and application-security teams who were responsible for securing products that legacy tools struggled to understand.
Funding an AI Offensive
The company's trajectory has been capital-efficient and steadily ambitious. It raised a $1 million pre-seed, followed by a $3.9 million seed round led by Iris in June 2023 [TechCrunch, June 2023]. The recent $18 million Series A in March 2026, led by Balderton Capital with participation from Uncorrelated Ventures, Iris, and Y Combinator, signals a scale-up phase [Tech.eu, March 2026]. The total disclosed funding now sits at approximately $22.9 million.
| Round | Date | Amount | Lead Investor |
|---|---|---|---|
| Pre-seed | 2023 | $1,000,000 | Frst [Startup Savant, October 2024] |
| Seed | June 2023 | $3,900,000 | Iris [TechCrunch, June 2023] |
| Series A | March 2026 | $18,000,000 | Balderton Capital [Tech.eu, March 2026] |
The proceeds are clearly earmarked for growth. The company has opened offices in New York, San Diego, Paris, and Amsterdam, and its headcount has grown from 10 employees at the seed announcement to more than 30 today [TechCrunch, June 2023] [Work at a Startup, March 2026]. Active hiring focuses on senior engineering roles in AI and security, a Head of Sales in New York, and customer success, indicating a push into the enterprise sales motion [Y Combinator].
Traction and the Target Customer
Escape reports it is now trusted by more than 100 customers [Built In, August 2026]. Its roster includes notable names like Zoom, Schibsted, Wiz, and PandaDoc, suggesting it can land and expand within tech-forward enterprises [Escape job posting]. For these customers, the platform promises to correlate code-to-cloud assets, manage the external attack surface, and conduct dynamic application security testing (DAST) through autonomous AI agents [Escape Careers].
In a recent internal analysis cited by the company, its technology identified over 2,000 high-impact vulnerabilities across live applications built with automated coding tools, including 175 cases involving exposure of personal data [Anna Nabiullina - Escape | LinkedIn]. This kind of result is the core value proposition: finding consequential flaws that other methods miss.
The ideal customer profile is an application-security or platform team at a company with a modern, API-driven product stack. These are organizations for whom manual pentests are a bottleneck and whose attack surface is too complex for conventional scanners. The budget owner likely sits in the CISO's organization, but the daily users are security engineers and developers who need actionable results, not just alerts.
The Realistic Competitive Set
Escape operates in a crowded security market, but its positioning is distinct. It is not directly competing with legacy SAST or DAST scanners that rely on signature databases. Instead, its realistic competitors are manual penetration testing services, bug-bounty platforms, and the emerging class of AI-powered security co-pilots. Its bet is that continuous, automated exploitation driven by AI agents will prove more efficient and effective than periodic human-led tests or crowdsourced programs. The sales cycle involves displacing a portion of an existing offensive security budget, not just adding another tool to the stack.
What to Watch in the Next Twelve Months
The next year will be about proving the platform's scale and stickiness. Key milestones will be landing more enterprise logos beyond the initial 100, demonstrating renewal and expansion rates at higher annual contract values, and continuing to evolve the AI agents' capability to handle increasingly complex application environments. The open Head of Sales role points to a concerted push into the U.S. market, where competition is fiercest but budgets are largest.
The risk, as with any AI-centric product, is in the execution. The technology must consistently find novel, high-severity vulnerabilities without overwhelming teams with false positives or, worse, causing disruptions in live environments. The company's answer, rooted in Carossio's background as a former pentester for French intelligence and a machine learning researcher at Apple, is that a deep understanding of both attack techniques and AI is foundational [Antoine Carossio | GraphQLConf 2023]. The $18 million Series A is a vote of confidence that they can build that bridge.
Sources
- [TechCrunch, June 2023] Escape dynamically scans APIs to find security flaws | https://techcrunch.com/2023/06/05/escape-dynamically-scans-apis-to-find-security-flaws/
- [Tech.eu, March 2026] Escape secures $18M Series A to develop AI cybersecurity agents | https://tech.eu/2026/03/10/escape-secures-18m-series-a-to-develop-ai-cybersecurity-agents/
- [Startup Savant, October 2024] Escape - Startup Company Profile | https://startupsavant.com/startup-center/escape
- [Built In, August 2026] Built In | https://builtin.com/job/sales-director-us-ai-appsec/10067864
- [Work at a Startup, March 2026] Work at a Startup | https://www.workatastartup.com/jobs/91356
- [Escape Careers] Escape Careers | https://escape.tech/careers
- [Y Combinator] Y Combinator | https://www.ycombinator.com/companies/escape/jobs
- [Anna Nabiullina - Escape | LinkedIn] Anna Nabiullina - Escape | LinkedIn | https://www.linkedin.com/in/anna-nabiullina/
- [Antoine Carossio | GraphQLConf 2023] Antoine Carossio | GraphQLConf 2023
- [Escape job posting] Escape job posting | https://jobs.ashbyhq.com/escape/32e02dde-b428-4ed5-8a50-51e98fca5ddc