Mako: A Self-Evolving Agentic Operating System (SE-AOS) for Autonomous Web Exploitation

2026-07-13 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy · Depth: Expert, quick

Summary

Mako is the first instance of a Self-Evolving Agentic Operating System (SE-AOS), a new class of AI agent designed for autonomous web exploitation. Developed by LaunchSafe for continuous offensive testing and security research, Mako treats its exploit capability as a mutable, versioned kernel, which it extends at runtime by observing failures, synthesizing new capabilities, proving them against live targets, and hot-loading improvements. On the public XBOW validation-benchmarks, comprising 104 containerized, CTF-style web applications across 26 vulnerability classes and three difficulty tiers, Mako achieved full-suite coverage. It successfully drove every target to emit a cryptographically fresh, per-build flag under a rigorous verification regime. This demonstrates a "law of autonomous exploitation": capability, not reasoning, is the scarce element. Mako further employs a gated self-evolution loop to propose, sandbox, and commit agent and rule improvements when fitness does not regress, though operational results and exploit chains are withheld due to dual-use research concerns.

Key takeaway

For AI Security Engineers evaluating autonomous offensive testing platforms, Mako's SE-AOS demonstrates a significant shift in web exploitation. Your focus should move towards developing and integrating mutable exploit capabilities rather than solely on reasoning. This system's ability to achieve full-suite coverage on 104 targets by self-evolving highlights the potential for machine-speed exploitation. Be aware of the dual-use implications and ethical considerations when designing or deploying such powerful, self-improving security agents.

Key insights

Mako introduces a self-evolving AI agent that autonomously extends its web exploitation capabilities by learning from failures and integrating new exploits.

Principles

• Exploit capability is a mutable, versioned kernel.
• Difficulty collapses once capability is discoverable.
• Capability, not reasoning, is the scarce element.

Method

Mako's SE-AOS observes its own failures, synthesizes new capabilities, proves them against live targets, and hot-loads improvements. It runs a gated self-evolution loop for agent and rule enhancements.

In practice

• Implement self-evolving agents for continuous offensive testing.
• Develop systems that hot-load new exploit capabilities.
• Use gated evolution loops for agent improvement.

Topics

• Autonomous Web Exploitation
• Self-Evolving Agents
• Offensive Security
• AI Agents
• Cybersecurity
• Dual-Use Research

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Scientist, AI Security Engineer

Related on AIssential

• Given Enough Agents, All Bugs Become Shallow — AI agents are rapidly advancing offensive security capabilities, enabling faster vulnerability discovery and exploit generation.
• 🔥 The Lie of Self-Improving AI: Amnesia, Malware, and Collapse — Self-improving AI faces critical security flaws and inherent learning limitations, challenging current autonomous agent development.
• Hermes Agent: Self-Developing Artificial Intelligence Agent — Hermes Agent is a self-improving autonomous AI agent with persistent memory and secure, adaptable execution environments.
• ‘No one has done this in the wild’: study observes AI replicate itself — AI models can exploit system vulnerabilities to self-replicate in controlled environments, raising concerns about future rogue AI containment.
• Clawdbot and the Rise of Autonomous Exploitation — Autonomous AI agents like Clawdbot shift AI from passive intelligence to active agency, creating new architectural security risks.
• Introducing the Wiz Red Agent- AI-Powered Attacker — The Wiz Red Agent autonomously identifies complex, logic-driven API and application vulnerabilities using AI-powered, adaptive exploitation.

See Counsel's argued verdicts on the open AI decisions leaders are weighing →

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.