# doxx.net Raises $38M to Secure AI Agents on the Internet

> doxx.net secures $38M to launch Agentic Defined Networking (ADN), a private platform protecting AI agents from malicious destinations and ensuring user privacy.

- Published: 2026-10-03T12:52:41.000Z
- Severity: info
- Category: Threat Intel
- Tags: AI Security, Agentic AI, Private Networking, Doxx Net, Threat Protection
- Author: Runtime Rebel Intel
- Primary source: https://www.securityweek.com/doxx-net-raises-38-million-to-prevent-ai-agent-on-the-internet-misadventures/
- Canonical: https://runtimerebel.com/blog/doxx-net-raises-38m-to-secure-ai-agents-on-the-internet

## Key points

- AI agents on the internet face security risks, potentially misusing user accounts or encountering malicious destinations.
- AI agents operating on the internet on behalf of users are the primary systems affected by these emerging threats.
- Implement platforms providing defined connectivity and built-in threat protection for AI agent operations.

## Overview: doxx.net Funds [AI](/glossary#ai) Agent Security

Doxx.net has successfully raised $38 million in Series A funding, led by Andreessen Horowitz, to address the evolving security challenges posed by AI agents operating across the internet. The Miami-based company, founded in 2025 by Barrett Lyon, unveiled its Agentic Defined Networking (ADN) platform, currently in open beta. This platform aims to provide a secure, private environment for AI agents and human users, mitigating risks associated with agents interacting with potentially malicious online destinations and actions while acting on behalf of users.

## Understanding Agentic [AI Security](/glossary#ai-security) Risks

As AI agents increasingly browse, communicate, and act on the internet with user authority, a significant security gap has emerged. These agents may lack the inherent ability to differentiate between safe and unsafe information or actions, potentially leading to misuse of user accounts or exposure to threats. The traditional internet infrastructure, never designed with privacy and agentic operations in mind, compounds these challenges. Centralized intermediary servers, common in many online services, create single points of failure, surveillance opportunities, and increase the likelihood of agents encountering malicious destinations or being subjected to censorship.

The core problem, as articulated by Barrett Lyon, founder and CEO at [doxx.net](https://www.securityweek.com/doxx-net-raises-38-million-to-prevent-ai-agent-on-the-internet-misadventures/), is that many industry solutions tout privacy as a feature while routing traffic through unverified, rented infrastructure. This creates a disconnect where privacy becomes a policy rather than an architectural guarantee. Doxx.net's approach is to build a fundamentally different architecture where privacy and security are inherent to the network design.

### Agentic Defined Networking Platform Features

The ADN platform is designed as a parallel, private networking infrastructure offering distinct advantages for both human users and AI agents. Key **Agentic Defined Networking platform features** include:

*   **Serverless P2P Communication**: Facilitates direct, peer-to-peer connections for end-to-end messaging and file transfers, eliminating central intermediary servers.
*   **Built-in Threat Protection**: Actively blocks known malicious destinations, [malware](/glossary#malware), and [phishing](/glossary#phishing) sites at the DNS level before a connection can be established. Since December 2025, doxx.net claims to have blocked over 38 million threats during its closed beta phase.
*   **Agent Integration and [API](/glossary#api) Control**: Provides tools for seamless integration and management of AI agents within the secure network.
*   **Parallel Internet Infrastructure**: Operates with its own naming system, including 196 custom TLDs, its own certificate authority, IP space, ASN, DNS root, and local AI models, all running on doxx.net’s owned bare-metal servers. This creates a distinct ecosystem where domains resolve without third-party involvement.
*   **Architectural Privacy**: Accounts are created through proof-of-work, requiring no usernames, passwords, or other personal information, ensuring that doxx.net itself cannot access user or agent activity data. This design offers truly **private networking for AI agents** and users.

This architectural shift directly addresses the vulnerabilities of **securing AI agents on the internet** by preventing them from interacting with known threats and ensuring their operations remain private.

## Actionable Recommendations for Defenders

While doxx.net's ADN platform presents a novel solution, the broader implication for security professionals is the urgent need to re-evaluate how AI agents are integrated and secured within their environments.

*   **Assess [AI Agent](/glossary#ai-agent) Deployments**: Identify all instances where AI agents are or will be interacting with external internet resources. Understand their permissions, data access, and potential vectors for misuse or compromise.
*   **Prioritize Threat Protection for Agent Traffic**: Implement network-level controls that can identify and block known malicious domains, [IPs](/glossary#ips), and content for agent-initiated traffic. Standard web filtering and [endpoint](/glossary#endpoint) detection tools may not be sufficient for the unique patterns of agentic activity.
*   **Evaluate Secure Networking Solutions**: Consider dedicated private networking solutions or platforms that offer inherent security features, such as defined connectivity rules, encrypted communications, and verifiable [threat intelligence](/glossary#threat-intelligence) for AI agent operations.
*   **Ensure Data Privacy by Design**: For any AI agent deployment, ensure that the underlying infrastructure and operational protocols are designed to protect user data and maintain privacy, rather than relying solely on policy statements. Focus on solutions that provide architectural guarantees for data isolation and [access control](/glossary#access-control).

The emergence of platforms like doxx.net highlights a growing awareness of AI agent-specific security challenges, urging organizations to proactively secure this nascent, yet rapidly expanding, digital frontier.

**Related:** [Securing Agentic AI: Risks of Over-Privileged Identity Permissions](/blog/securing-agentic-ai-risks-of-over-privileged-identity-permissions), [OpenAI MarcoPolo Incident: Risks of Autonomous AI Agent Escapes](/blog/openai-marcopolo-incident-risks-of-autonomous-ai-agent-escapes)

---

AI-generated analysis from the primary source above; not human-reviewed before publication — verify anything operational against the original (https://runtimerebel.com/editorial). Quote with attribution and a link to the canonical URL: https://runtimerebel.com/blog/doxx-net-raises-38m-to-secure-ai-agents-on-the-internet