XBOW Interview Guide (2026): Autonomous Hacking, Security Depth, and AI Agents

XBOW is an autonomous offensive security company building an AI system that finds and exploits software vulnerabilities without a human driving each step. It was founded by Oege de Moor, the Oxford professor turned entrepreneur who created Semmle and its CodeQL analysis engine, which GitHub acquired and where he went on to lead GitHub Advanced Security and work on Copilot; much of the early team came from that GitHub group. The product runs many AI agents in parallel to probe web and application targets, and in 2025 it became the first machine to reach #1 on HackerOne’s global leaderboard, reporting more than a thousand validated vulnerabilities. The company is based in Seattle and runs remote-first. It is well funded, with a $75M Series B in 2025 led by Altimeter alongside Sequoia and Nat Friedman, followed by a $120M Series C in March 2026 led by DFJ Growth and Northzone at a valuation above $1B.

Process

Public detail on XBOW’s exact loop is thin, so treat this as the likely shape based on stage and domain rather than a fixed script. Expect a recruiter screen first (background, why offensive security, whether a small remote-first startup fits you), then a technical screen, then an onsite of roughly three to five rounds. For security roles, expect hands-on vulnerability work: read a real web app, find the bug, and explain how you would exploit it. For engineering roles, expect strong coding, a system design round tied to their agent platform or evaluation harness, and a behavioral conversation with a founder or lead about ownership and moving fast. A company this small can compress the whole thing into one to two weeks when they like you, so keep your scheduling flexible.

What they actually ask

• Web and application vulnerabilities: walk through classes like SQL injection, SSRF, auth bypass, IDOR, and XSS, and how you would find and confirm each one on a live target.
• Exploitation depth: given a suspected bug, how you build a working proof of concept, chain steps together, and rule out false positives.
• LLM-agent design for security tasks: how you would structure agents that plan, call tools, recover from dead ends, and stay on task across long autonomous runs.
• Program analysis and tooling: static and dynamic analysis, fuzzing, and taint tracking, plus how you would feed that signal into an agent (a nod to the company’s CodeQL heritage).
• Sandboxing and safety: running untrusted code and attacker payloads in isolation, and keeping an autonomous system from causing damage outside its scope.
• Evaluating an autonomous hacker: how you measure whether the agent is actually good, including coverage, false-positive rate, and how you avoid overfitting to a single leaderboard. Strong, clean coding under time sits underneath all of it, and the real bar is the mix of security depth and AI-agent engineering.

Levels and comp (2026)

Compensation is base plus equity, with equity granted as options or RSUs in a private company on a four-year vest and usually a one-year cliff. XBOW is new and small enough that it may not carry its own levels.fyi bands yet, so anchor on comparable well-funded security-AI startups in Seattle and the Bay Area rather than a single quoted number. Based on that market, mid-career engineers tend to land roughly in the $180K to $300K total range, with senior and staff levels and specialized security research running higher; base salaries commonly sit in the $160K to $230K band, and equity makes up the rest. These bands shift with location and level, so check levels.fyi and recent Blind threads for current numbers. Be clear-eyed about the equity: it carries real upside at a company raising fast above a $1B valuation, but it is illiquid until an exit, and the unit count and strike or preferred price matter more than the headline figure a recruiter quotes. Ask for the base, the number of units, the current 409A or preferred price, and the vesting terms so you can value the offer yourself.

Prep priorities

Split your prep between security depth and agent engineering, because the bar sits at the intersection of the two. Get sharp on real web and app exploitation: practice on hands-on labs, read recent public writeups and CVEs, and be ready to explain a full exploit chain instead of just naming a bug class. If you come from ML or backend, learn how autonomous agents are built and where they break, including planning, tool use, context limits, recovery, and evaluation. Know program analysis basics such as static analysis, fuzzing, and taint tracking, since that lineage runs straight through the company. Be ready to reason about sandboxing and running untrusted payloads safely. Bring two or three project stories where you owned something end to end and shipped under ambiguity. And write correct, readable code while narrating your thinking, since interviewers grade how you reason as much as the final answer.

Frequently Asked Questions

Do I need offensive security experience to get hired? It helps a lot for security-facing roles, but it is not required for every seat. XBOW also hires ML, agent, and backend engineers, so strong fundamentals plus real curiosity about how systems break can carry you, as long as you are ready to pick up the security side fast.

Is the interview LeetCode-heavy? Expect practical coding over obscure puzzles based on the company’s stage and domain, but you should still be fluent with core data structures and able to write correct code quickly. The harder signal is your security reasoning and how you think about agent design.

Is XBOW remote? Yes. It runs remote-first from a Seattle base, so expect a video loop rather than an onsite. Confirm time-zone overlap and any travel expectations with your recruiter early.

How should I talk about their HackerOne result? Know the story (first machine to reach #1, more than a thousand reported bugs) but focus on the how: what makes an autonomous hacker reliable, how you keep false positives low, and where you would push the system next.