XBOW arrived in January 2024 and, by mid-2025, had become the first autonomous system to top HackerOne’s US leaderboard by reputation. Founded by Oege de Moor, creator of GitHub Copilot, it has since raised around $237 million and sells largely per engagement, with on-demand pentests starting in the low thousands. It is a genuinely impressive product — but it is web-application-shaped, US-hosted, priced per test rather than continuously, and its API and mobile coverage is still maturing. If any of those constraints matter to you, the five platforms below approach autonomous offensive security from meaningfully different angles.
1. Escape
- Founding year: 2020
- Overview: Escape is the strongest alternative for teams whose real exposure sits in APIs and application business logic rather than classic web vulnerabilities. Founded in Paris in 2020 by Tristan Kalos and Antoine Carossio, it raised an $18M Series A led by Balderton in March 2026 and now positions itself as an offensive security engineering platform. Its Cascade engine runs continuous agentic pentests that chain multi-step attacks and return proof of exploit, including white-box source-code testing, while its business-logic DAST handles stateful multi-user and multi-tenant flows to catch BOLA, IDOR and broken access control natively across REST, GraphQL and gRPC. Findings convert into automated regression tests in under an hour and ship with framework-specific code fixes. Over 2,000 security teams use it, with EU and US data residency available.
- Services provided: Attack surface management; business-logic-aware DAST; AI pentesting; external network pentesting; GraphQL and gRPC security testing; AI false-positive triage; automated regression test generation; CI/CD, Jira and Slack integrations; SOC 2, ISO 27001 and PCI-DSS compliance reporting.
- Web address: https://escape.tech/
2. Horizon3.ai
- Founding year: 2019
- Overview: Horizon3.ai is the volume leader in autonomous pentesting, with NodeZero deployed at more than 6,500 organisations and over 300,000 production pentests run. It was founded in 2019 by former Splunk CTO Snehal Antani and former USSOCOM deputy CTO Anthony Pillitiere, and raised a $250M Series E in August 2026 at a valuation above $2 billion. Where XBOW is application-first, NodeZero is infrastructure-first: it chains weak credentials, misconfigurations and identity gaps into real attack paths across internal networks, Active Directory, cloud, hybrid environments and Kubernetes, then re-tests automatically to confirm fixes landed. It deploys as a run-once, unauthenticated Docker container with no persistent agents or provisioned credentials, which suits sensitive and air-gapped networks. Web application pentesting arrived in July 2026 and is the newest part of the portfolio.
- Services provided: Internal and external network pentesting; cloud and Kubernetes pentesting; credential and identity attack path testing; web application pentesting; automated fix verification; decoy seeding; MSP and channel delivery.
- Web address: https://horizon3.ai/
3. Pentera
- Founding year: 2015
- Overview: Pentera defined the automated security validation category before AI-native offensive security existed, launching in Tel Aviv as Pcysys in 2015 and rebranding in 2021. It has raised roughly $250 million from Insight Partners, K1, Evolution Equity and Blackstone, passed $100 million in ARR and serves over 1,100 enterprises. The platform is agentless and modular: Core for internal network testing, Surface for external attack surface, Cloud for hybrid attack paths, plus credential exposure and ransomware readiness modules. Its distinguishing question is not only whether something is exploitable but whether your existing controls stopped it — validating the security stack itself, which suits mature CTEM programmes and regulated buyers. It can be self-hosted for internal validation. Application logic depth is thinner than XBOW’s, and its AI layer orchestrates rather than drives the engine.
- Services provided: Internal network pentesting; external attack surface validation; cloud and hybrid attack path testing; credential exposure testing; ransomware readiness validation; security control validation; agentic AI attack execution and attack path analysis.
- Web address: https://pentera.io/
4. RunSybil
- Founding year: 2023
- Overview: RunSybil is the closest philosophical twin to XBOW and the most credible option if you want AI-native pentesting without handing over source code. Founded in 2023 by Ari Herbert-Voss, OpenAI’s first security hire, and Vlad Ionescu, formerly head of offensive security red teaming at Meta, it raised $40 million in March 2026 led by Khosla Ventures with participation from the Anthology Fund backed by Anthropic and Menlo Ventures. Sybil runs continuous black-box testing against live applications and infrastructure with no code access — finding forgotten endpoints, probing authentication boundaries, chaining vulnerabilities and demonstrating real data exposure, then retesting within hours. It reports over 90% fewer false positives than conventional scanners. Customers include Cursor, Notion, Baseten and Thinking Machines Lab, on a subscription rather than per-pentest model.
- Services provided: Continuous autonomous black-box pentesting; attack surface and endpoint discovery; authentication and authorisation boundary testing; vulnerability chaining and live exploitation; rapid retesting; false-positive suppression.
- Web address: https://www.runsybil.com/
5. Terra Security
- Founding year: 2024
- Overview: Terra Security is the alternative for buyers who are not comfortable with full autonomy. Founded in 2024 by Shahar Peled and Gal Malachi with offices in New York and Tel Aviv, it has raised $38 million including a $30 million Series A led by Felicis, and counts former Google CISO Gerhard Eschelbeck on its board. The platform runs a swarm of agentic AI pentesters with a human expert on the loop validating every finding — slower than fully autonomous testing, but considerably easier to defend to an auditor or a board. Coverage is broader than XBOW’s, spanning web applications, external and internal networks and AI systems, including red teaming of copilots, LLMs and MCP servers. It holds SOC 2 Type II and CREST membership, and compresses multi-week manual engagements into hours.
- Services provided: Agentic AI web application pentesting; external and internal network pentesting; AI and LLM red teaming; MCP server testing; human-validated exploitation; business-context risk scoring; SOC 2, ISO 27001 and HIPAA-ready reporting.
- Web address: https://www.terra.security/
