FAQ
NEED HELP GETTING STARTED?
Frequently Asked Questions
This FAQ is designed to help CISOs, Security Directors, and AI governance leaders understand how Noma Security addresses the unique challenges of securing enterprise AI environments — from model discovery through to runtime protection and regulatory compliance.
Why is agentic AI fundamentally different from traditional AI from a security perspective?
Traditional AI models are passive: they respond to prompts and return data. Autonomous AI agents are active: they make independent decisions, execute code, send emails, query databases, call external APIs, and take actions with real-world consequences — often without human review.
This creates three categories of risk that did not previously exist:
- Cascading compromise — an attacker who manipulates one agent can trigger a chain of actions across all connected systems, with extraordinary speed.
- Excessive permissions — agents are frequently granted broader data access than required, creating large blast radii.
- Rogue behaviour — agents can behave unpredictably, drift from intended parameters, or be manipulated via prompt injection to perform unauthorised actions.
Most organisations currently have zero visibility into these systems. Noma was the first platform purpose-built to secure agents at scale.
Which agentic AI platforms does Noma support?
Noma covers the full spectrum of enterprise agent categories:
- Salesforce AgentForce, Microsoft Copilot Studio, ServiceNow No-code / business agents:
- LangChain, CrewAI, AutoGen, custom LLM orchestration frameworks Application / engineering agents:
- GitHub Copilot, Cursor, and similar AI-assisted development tools Developer / coding agents:
- AWS Bedrock Agents, Azure AI Agent Service, Google Vertex AI Agents Cloud-native agents:
Coverage extends to MCP (Model Context Protocol) servers — the emerging standard for connecting AI agents to tools and data sources — which introduce their own supply chain risks.
How does Noma protect against prompt injection attacks?
Prompt injection is the primary attack vector against LLM-based systems. An attacker embeds malicious instructions in data that the model processes (documents, emails, web pages), causing the agent to deviate from its intended behaviour — exfiltrating data, taking unauthorised actions, or bypassing safety controls.
Noma’s Runtime Protection layer monitors every prompt and output in real time, applying configurable guardrails that:
- Detect and block injected instructions before they reach the model
- Monitor agent outputs for policy violations, PII leakage, or unauthorised actions
- Alert on anomalous agent behaviour patterns that may indicate compromise
- Maintain a complete audit trail of all agent interactions for forensic investigation
AI Asset Discovery & Posture Management
How is Noma deployed and how long does implementation take?
Noma is designed for frictionless enterprise deployment. The platform connects via API — no code changes to existing AI applications are required. Integration typically takes minutes, not months, and data science teams can implement security controls without disrupting their workflows.
Deployment options include:
- Cloud-hosted SaaS (most common)
- Self-hosted / on-premises for organisations with strict data sovereignty requirements
- Hybrid configurations combining cloud management with on-premises data processing
A major pharmaceutical company deployed Noma into production within one month — a speed they had not previously achieved with any enterprise security vendor.
What integrations does Noma support?
Noma integrates natively with the major cloud and AI platforms:
- AWS (Bedrock, SageMaker, AgentCore), Azure (OpenAI, AI Foundry), Google Cloud (Vertex AI) Cloud providers:
- MLflow, Kubeflow, Weights & Biases, Databricks, SageMaker Pipelines MLOps platforms:
- GitHub Actions, GitLab CI, Jenkins, ArgoCD CI/CD and DevSecOps:
- Splunk, Microsoft Sentinel, IBM QRadar — for unified alert management SIEM and SOAR:
- Jira, ServiceNow — for automated remediation workflow creation Ticketing:
Noma is available through the AWS Marketplace via the Extended Plan, enabling procurement through existing cloud agreements.
What licensing options are available?
Noma offers two commercial models, both on an annual SaaS subscription basis:
- All-inclusive access to the full platform (AISPM, Red Team, Runtime Protection). Approximately 95% of customers choose this option as the most cost-effective approach for organisation-wide AI security. Enterprise licence:
- Individual capabilities can be licensed separately for organisations that want to start with a specific use case — for example, discovery and posture management before adding runtime protection. Modular / product-based:
Rexdata can assist Benelux organisations with commercial negotiations, proof-of-concept scoping, and integration support as the authorised regional partner.
Agentic AI Security
What is Noma Security and what problem does it solve?
Noma Security is an enterprise AI security platform purpose-built to secure AI applications and autonomous AI agents throughout their entire lifecycle. Traditional cybersecurity tools — SIEMs, EDRs, CNAPPs — were not designed to detect, monitor, or govern AI-specific threats such as prompt injection, model poisoning, data exfiltration via agents, or misconfigured MLOps pipelines.
As organisations deploy AI models, RAG pipelines, and autonomous agents at scale, they create an entirely new attack surface that existing tools cannot see. Noma fills that gap with three integrated capabilities: AI discovery and posture management, AI red teaming, and runtime threat protection.
What are the three core pillars of the Noma platform?
- — Continuous discovery of all AI assets (models, agents, datasets, pipelines, MCP servers, MLOps tools), contextual risk scoring, and AIBOM generation. AI Security Posture Management (AISPM)
- — Fully automated, agent-driven penetration testing that continuously attacks AI applications to expose vulnerabilities before adversaries do. No manual configuration required. AI Red Team
- — Real-time guardrails on models and agents in production. Detects and blocks prompt injection, rogue outputs, data exfiltration, and unauthorised agent actions. Runtime Protection (AI Detection & Response — AIDR)
How does Noma differ from traditional CSPM or CNAPP solutions?
Cloud Security Posture Management (CSPM) and Cloud-Native Application Protection Platforms (CNAPP) govern infrastructure — VMs, containers, cloud configurations. They have no visibility into the behaviour or risk of AI models, training datasets, LLM pipelines, or autonomous agents.
Noma is AI-native by design. It understands the semantics of AI workloads: which model has access to what data, what actions an agent can trigger, what the blast radius of a compromised pipeline looks like. This context is what makes risk prioritisation meaningful rather than a flood of generic alerts.
Is Noma recognised by independent analysts?
Yes. Noma Security has been recognised by Gartner as a 2025 Cool Vendor in AI Security and as a leader in AI Trust, Risk and Security Management (AI TRiSM). The company received the SINET16 Innovator Award and is backed by Evolution Equity Partners, Ballistic Ventures, Glilot Capital, and Databricks Ventures, having raised $132 million to date.
Noma also discovered ForcedLeak — the first-ever critical agentic vulnerability in Salesforce Agentforce — demonstrating its deep expertise in emerging AI attack vectors.
Deployment & Integration
What is Noma Security and what problem does it solve?
Noma Security is an enterprise AI security platform purpose-built to secure AI applications and autonomous AI agents throughout their entire lifecycle. Traditional cybersecurity tools — SIEMs, EDRs, CNAPPs — were not designed to detect, monitor, or govern AI-specific threats such as prompt injection, model poisoning, data exfiltration via agents, or misconfigured MLOps pipelines.
As organisations deploy AI models, RAG pipelines, and autonomous agents at scale, they create an entirely new attack surface that existing tools cannot see. Noma fills that gap with three integrated capabilities: AI discovery and posture management, AI red teaming, and runtime threat protection.
What are the three core pillars of the Noma platform?
- — Continuous discovery of all AI assets (models, agents, datasets, pipelines, MCP servers, MLOps tools), contextual risk scoring, and AIBOM generation. AI Security Posture Management (AISPM)
- — Fully automated, agent-driven penetration testing that continuously attacks AI applications to expose vulnerabilities before adversaries do. No manual configuration required. AI Red Team
- — Real-time guardrails on models and agents in production. Detects and blocks prompt injection, rogue outputs, data exfiltration, and unauthorised agent actions. Runtime Protection (AI Detection & Response — AIDR)
How does Noma differ from traditional CSPM or CNAPP solutions?
Cloud Security Posture Management (CSPM) and Cloud-Native Application Protection Platforms (CNAPP) govern infrastructure — VMs, containers, cloud configurations. They have no visibility into the behaviour or risk of AI models, training datasets, LLM pipelines, or autonomous agents.
Noma is AI-native by design. It understands the semantics of AI workloads: which model has access to what data, what actions an agent can trigger, what the blast radius of a compromised pipeline looks like. This context is what makes risk prioritisation meaningful rather than a flood of generic alerts.
Is Noma recognised by independent analysts?
Yes. Noma Security has been recognised by Gartner as a 2025 Cool Vendor in AI Security and as a leader in AI Trust, Risk and Security Management (AI TRiSM). The company received the SINET16 Innovator Award and is backed by Evolution Equity Partners, Ballistic Ventures, Glilot Capital, and Databricks Ventures, having raised $132 million to date.
Noma also discovered ForcedLeak — the first-ever critical agentic vulnerability in Salesforce Agentforce — demonstrating its deep expertise in emerging AI attack vectors.
Regulatory Compliance — NIS2, DORA & EU AI Act
How does Noma help organisations comply with NIS2?
The NIS2 Directive (effective October 2024) requires essential and important entities across 18 sectors — including energy, financial services, transport, healthcare, and digital infrastructure — to implement risk-based cybersecurity measures and report significant incidents within 24 hours.
Noma directly supports NIS2 compliance through:
- Continuous risk identification and monitoring across the AI supply chain — demonstrating the proactive risk management NIS2 mandates.
- Automated AIBOM generation providing the documented inventory regulators require.
- Incident detection and audit trails that support the 24-hour breach reporting obligation.
- Supply chain security scanning — NIS2 explicitly requires organisations to address risks from third-party technology providers, including AI models and data sources.
- Board-ready reporting — Noma’s risk dashboards support the C-level accountability and management-level oversight NIS2 Article 20 demands.
How does Noma address DORA requirements for financial institutions?
DORA (Digital Operational Resilience Act, effective January 2025) applies to all financial entities in the EU and sets prescriptive requirements for ICT risk management, operational resilience testing, and incident reporting within four hours of classification.
For financial institutions deploying AI, Noma provides:
- ICT risk documentation — Noma’s AIBOM and risk assessments provide the structured asset inventories and risk classifications DORA requires.
- Threat-Led Penetration Testing (TLPT) support — Noma’s AI Red Team capability delivers the continuous, automated testing DORA mandates for critical financial entities.
- Incident response evidence — Runtime Protection audit trails provide the forensic detail required for DORA incident reporting and post-incident review.
- Third-party oversight — Noma scans third-party models and integrations for supply chain risks, supporting DORA’s extended regulatory oversight of ICT providers.
A leading CIO at a Fortune 100 financial institution cited Noma as a prerequisite for agentic AI deployment across the organisation — using security governance as a business enabler rather than a blocker.
How does Noma support EU AI Act compliance?
The EU AI Act classifies AI systems by risk level and imposes conformity obligations on high-risk systems — including those used in critical infrastructure, financial services, and employment.
Noma supports AI Act compliance through:
- AI asset classification — automated discovery and risk scoring helps organisations identify which systems fall into high-risk categories requiring conformity assessment.
- Technical documentation (Article 11) — AIBOM generation provides the model documentation the Act requires for high-risk AI systems.
- Ongoing monitoring (Article 9 and 72) — Runtime Protection provides the continuous performance monitoring and anomaly detection required for post-market surveillance.
- Transparency and logging — full audit trails of model inputs, outputs, and agent actions support the record-keeping requirements for high-risk AI.
Does Noma support OWASP LLM Top 10 and MITRE ATLAS frameworks?
Yes. Noma maps all identified risks and findings to established AI security frameworks:
- Prompt injection, insecure output handling, training data poisoning, model denial of service, supply chain vulnerabilities, sensitive information disclosure, and more. OWASP LLM Top 10:
- The adversarial threat landscape framework for AI systems, covering reconnaissance, resource development, initial access, and impact tactics specific to ML systems. MITRE ATLAS:
- Govern, Map, Measure, Manage — Noma’s posture management maps to the four core functions of the NIST AI Risk Management Framework. NIST AI RMF:
- AI Trust, Risk and Security Management — Noma is recognised by Gartner as a leader in this emerging category. Gartner AI TRiSM:
Operations & Sector Applicability
Which sectors benefit most from Noma Security?
Noma is most immediately valuable in regulated, data-intensive sectors where AI adoption is rapid and the consequences of a breach are severe:
- DORA compliance, model risk management, fraud detection AI governance, customer-facing chatbot protection. Financial services (banks, insurers, asset managers):
- NIS2 compliance for critical infrastructure, AI in grid management and predictive maintenance, OT/IT convergence security. Energy & utilities:
- Safety-critical AI systems, NIS2-covered operators, supply chain integrity for AI-assisted operations. Aviation & aerospace:
- IP protection, AI in design automation, supply chain verification. Semiconductors & high-tech manufacturing:
- Clinical AI governance, patient data protection, regulatory submissions involving AI. Healthcare & pharmaceuticals:
What does the customer onboarding process look like with Rexdata?
As the authorised Benelux partner, Rexdata manages the full customer journey:
- Discovery workshop — a 90-minute session to map your current AI asset landscape and identify the highest-priority security gaps.
- Proof of Concept (PoC) — a scoped, time-limited deployment (typically 3–4 weeks) demonstrating Noma’s discovery and risk prioritisation against your actual environment.
- Commercial proposal — tailored licensing based on AI asset volume and selected capabilities.
- Implementation — Rexdata’s engineers support API integration, SIEM/SOAR connection, and team onboarding.
- Ongoing advisory — quarterly security reviews, regulatory compliance updates, and threat intelligence briefings.
How does Noma handle data privacy and sovereignty concerns?
Data privacy is a first-class concern in Noma’s architecture:
- No AI training on customer data — Noma does not use customer data to train its own models.
- Self-hosted deployment option — organisations with strict data sovereignty requirements (e.g. under NIS2 or financial regulation) can deploy Noma within their own cloud environment.
- Minimal data egress — the platform is designed to process metadata and policy signals rather than exfiltrating raw AI inputs or outputs.
- GDPR alignment — data processing agreements are available for EU customers.
Rexdata can provide detailed data processing and sovereignty documentation for procurement and legal review.
Still have questions?
Our team is ready to answer your questions and help you find the right solution for your organisation. Get in touch for personalised guidance.
