UMAI is Now Generally Available
Today we are excited to announce that UMAI is generally available. Any enterprise team can now sign up and start governing their AI systems in minutes.
This moment has been six months in the making. In June 2024 we launched UMAI's private beta with 12 enterprise teams. By November we had 40 teams across financial services, healthcare, insurance, government, and manufacturing actively using UMAI to govern their AI deployments.
Here is what we built, what we learned, and where we are headed.
What We Built
UMAI is the control plane for enterprise AI. It sits inline between your applications and AI models, providing:
PRE_LLM Enforcement. Every prompt passes through UMAI before reaching the model. We detect and block prompt injection, data leakage, PII exposure, and policy violations in real time.
POST_LLM Enforcement. Every model response passes through UMAI before reaching the user. We detect hallucinated PII, toxic content, off-topic responses, and compliance violations.
Browser AI Governance. UMAI's browser extension governs employee interactions with ChatGPT, Claude, Gemini, and other AI tools — preventing sensitive data from being pasted into consumer AI products.
Tamper-Evident Audit Trail. Every request, response, policy evaluation, and enforcement action is recorded in a hash-chained audit log that provides cryptographic proof of governance.
Sovereign Deployment. UMAI can run entirely on your infrastructure — on-premises, private cloud, or air-gapped environments — with no data leaving your network.
What We Learned in Beta
Latency is Non-Negotiable
Our beta customers told us clearly: if your guardrail adds more than a few milliseconds of latency, we will turn it off. We took this seriously. UMAI's median added latency is 1.8ms, with p99 under 4ms. We achieved this through aggressive optimization of our detection pipeline, edge deployment, and a fail-open option for customers who prioritize availability over enforcement.
Policy Flexibility Matters More Than Policy Coverage
Early in the beta we focused on adding more detection capabilities — more PII types, more injection patterns, more content categories. Our customers told us coverage was important, but what they really needed was flexibility: the ability to define custom policies, set per-team enforcement levels, and create exception workflows.
We responded by building UMAI's policy engine as a fully programmable rule system. Customers can define policies using our visual editor or write them as code (YAML or JSON). Policies support conditions, exceptions, severity levels, and custom actions.
Compliance Teams Are the Primary Buyer
We expected security teams to be our primary audience. They are important, but the teams that drove procurement and deployment were compliance and risk teams — the people responsible for demonstrating AI governance to regulators, auditors, and board members.
This insight shaped our product: we invested heavily in the audit trail, compliance dashboards, and evidence export features that compliance teams need. The tamper-evident hash chain was initially a nice-to-have; it became our most requested feature.
GA Launch Features
In addition to everything from the beta, GA includes:
- Organization-scope policies with inheritance to projects and environments
- SSO and SCIM for enterprise identity management
- Slack and Teams integrations for real-time alerts
- Terraform provider for infrastructure-as-code deployment
- SOC 2 Type II compliance — UMAI itself is SOC 2 certified
- 99.99% uptime SLA on the Enterprise plan
What is Next
We are just getting started. Here is what is on our roadmap for the first half of 2025:
Federated Learning. Privacy-preserving model updates that improve detection accuracy across customers without sharing any customer data.
Agent Governance. As AI agents become more autonomous — executing code, making API calls, managing infrastructure — the governance requirements become more complex. UMAI will support policy enforcement for agentic workflows, including tool-use approval, action logging, and human-in-the-loop gates.
EU AI Act Compliance Toolkit. Purpose-built features for Article 9 (Risk Management), Article 13 (Transparency), and Article 14 (Human Oversight) compliance.
Get Started
UMAI is available today with a free tier for small teams and usage-based pricing for enterprises. Sign up at app.umaisolutions.com/signup or contact our sales team at contact@umaisolutions.com for a guided deployment.
Thank you to every beta customer who helped us build UMAI into what it is today. We are honored to be your AI governance platform.