doxx.net Secures $38 Million to Combat AI Agent Misadventures on the Internet

Post Views: 9

doxx.net has secured $38 million in Series A funding, led by Andreessen Horowitz with additional support from Animo Ventures and Focal.vc. The company, established in 2025 by founder Barrett Lyon, operates across six continents and unveiled its Agentic Defined Networking (ADN) platform during the same funding announcement. This platform, currently in open beta, provides a secure environment for AI agents to perform tasks while maintaining privacy and threat protection. The ADN architecture addresses vulnerabilities in traditional networking models by eliminating reliance on centralized servers. Lyon emphasized that existing solutions often prioritize profit over security, routing traffic through unverified infrastructure. In contrast, ADN’s design ensures privacy is embedded in the network’s structure rather than relying on policy documents. Users and agents control their own privacy parameters, with the platform explicitly avoiding knowledge of user activity, such as search history, visited websites, or communication partners.

Key Features of ADN

Key features of ADN include serverless peer-to-peer (P2P) communication, end-to-end encryption for messaging and file transfers, and integrated threat protection that blocks malicious domains, malware, and phishing sites. The platform supports AI agents with API controls and operates on a parallel internet infrastructure. It also includes DNS-level threat detection, preventing connections to compromised or emerging malicious sources. Since December 2025, the closed beta phase has blocked over 38 million threats.

Technical Specifications

ADN enables customers to deploy applications on a private network using its proprietary naming system, which includes 196 custom top-level domains (TLDs) and a self-managed certificate authority. Domains resolved through ADN are inaccessible via the public internet, ensuring no third-party involvement. The platform’s infrastructure includes dedicated IP space, autonomous system numbers (ASNs), a root DNS system, and localized AI models hosted on bare-metal servers. User authentication bypasses traditional credentials, relying instead on proof-of-work mechanisms to create accounts.

Quote from Andreessen Horowitz

Joel De La Garza, a partner at Andreessen Horowitz and board member at doxx.net, highlighted the platform’s unique approach to privacy. He stated that ADN offers a “one-stop privacy stack” built from first principles, enabling secure communications and operations for both humans and AI agents. The system’s design prioritizes privacy through architectural integrity rather than marketing claims.

Platform Capabilities

The ADN platform’s technical specifications include native support for iOS, macOS, Android, Windows, and Linux, all integrated into a global backbone network. Its threat mitigation capabilities extend to real-time blocking of malicious activity, while its decentralized framework reduces risks associated with censorship, surveillance, and single points of failure. The platform’s emphasis on user control and transparency aims to redefine privacy standards in an era of increasing AI adoption and cyber threats.