ZioSec

ZioSec is a cybersecurity company that builds an autonomous offensive-security ("red team") platform for AI agents and multi-agent systems. The platform fingerprints each target agent — its model, tools, memory and connections — then generates and executes a bespoke, deep-chained attack tree against it, chaining prompt injection into tool misuse, privilege escalation and data exfiltration rather than firing single-shot prompts. Testing is continuous: campaigns re-run on a schedule and whenever a model, prompt or tool changes. Destructive actions are simulated or approval-gated, with rate limits, a one-click stop and a full audit log so the platform can be pointed at production agents.

Every finding is recorded with severity, deterministic reproduction steps, remediation guidance covering immediate and long-term fixes, and a classification in ZioSec's own A2OSF taxonomy, described as 8 attack layers, 10 tactics and 62 techniques. That internal record is exported to six external frameworks — OWASP AISVS, MITRE ATLAS, NIST AI RMF, ISO 42001, the EU AI Act and AIUC-1 — so results can be used as audit evidence. The product also presents fleet-level views: per-agent risk, top risky attacks and attack-path coverage. ZioSec says it can test any agent that runs a model and calls tools, including custom agents, Claude Code, and agents exposed over MCP or A2A, with no agent SDK required.

ZioSec positions itself against adjacent categories it says do not attack a live agent: runtime guardrails that block known-bad inputs inline, evaluation tools that score model quality against fixed test sets, and model/supply-chain scanners.

Founding story

ZioSec was founded by security veterans and engineers who identified a gap in which AI agents were shipping faster than the tooling needed to test them. Co-founder and CTO Andrius Useckas previously founded ThreatX, an API and application security company acquired by A10 Networks, and has over 25 years of experience as a pentester; co-founder and CEO Aaron Walls is a Techstars and Cornell alumnus and multi-time founder.

Business model

ZioSec sells its offensive testing engine through three delivery modes: a self-serve platform for continuous validation across an agent fleet (inventory, campaigns, per-agent policy, executive risk posture); an API that accepts an agent endpoint and returns audit-ready evidence, aimed at TPRM, insurance and security platforms; and fixed-scope, expert-led pentest engagements.

Platform subscriptions and API access, plus fixed-scope expert-led pentest engagements priced from $10,000, which are credited 100% toward a platform subscription if the customer continues.

Traction

ZioSec reported running pilots with large enterprises and AI-forward security teams as of July 2025. Its site cites a finding-density dataset of n = 2,318 findings across attack layers and tactics, and lists a five-person team of security veterans and engineers.

Latest developments

The $2.1 million seed round was announced in July 2025, with proceeds directed at product development, go-to-market and expanding offensive testing coverage; the company markets platform, API and scoped-pentest delivery options alongside its A2OSF taxonomy and six-framework evidence exports.