doxx.net, a Miami, FL-based security startup, has raised $38 million in a Series A round led by Andreessen Horowitz, with participation from Animo Ventures and Focal.vc. The company was founded in 2025 by Barrett Lyon and operates across six continents.
The funding announcement coincided with the open beta launch of doxx.net’s Agentic Defined Networking (ADN) platform, built to address a growing gap in agentic AI security: autonomous agents increasingly browse the internet, communicate, and take action on a user’s behalf, but often cannot reliably distinguish safe destinations and actions from malicious ones while operating with the user’s credentials and authority.
How ADN Works
ADN is designed as a parallel, private network for both humans and AI agents, using serverless peer-to-peer communication rather than routing traffic through centralized intermediary servers. According to doxx.net, centralized infrastructure of this kind creates a single point for interception and surveillance, offers privacy guarantees that are more policy than architecture, raises the risk of agents reaching malicious destinations, and remains vulnerable to censorship and blocking.
The platform includes end-to-end encrypted messaging and file transfer, built-in threat protection against malicious destinations, malware, and phishing sites, API-level controls for agent integration, and anti-tracking features. doxx.net operates its own IP space and ASN, DNS root, certificate authority, bare-metal servers, and local AI models, along with 196 custom top-level domains that resolve only within its private network. Native apps are available for iOS, macOS, and Android, with Windows and Linux support on the company’s global backbone.
Accounts require no usernames, passwords, or personal information, instead relying on proof-of-work for account creation. The company says it has blocked more than 38 million threats during closed beta testing since December 2025.
Positioning Against Centralized Privacy Tools
Founder and CEO Barrett Lyon framed ADN as a response to what he described as an industry that markets privacy as a feature while routing traffic through third-party infrastructure that cannot be independently verified. He said doxx.net’s architecture is intended to ensure that users and agents, not operators, control privacy terms, and that the network itself cannot see search activity, browsing destinations, or communications.
Joel De La Garza, a partner at Andreessen Horowitz and a doxx.net board member, said the platform delivers privacy and security built from first principles rather than policy commitments, for use by both humans and AI agents.
The round adds doxx.net to a growing list of startups raising capital this year to secure agentic AI deployments, reflecting broader industry concern that autonomous agents acting with user-level authority represent a largely unaddressed attack surface.