Enterprise AI Is Moving Faster Than
Security Can Follow.
Security teams are being asked to govern AI systems, agents, providers, and workflows that move faster than traditional review cycles. The hard part is not only seeing AI risk. It is controlling it and proving the control worked.
Traditional URL filters were built
for human browsing.
A page can look completely harmless to a user while hiding instructions in CSS, comments, metadata, Unicode tags, or zero-width characters. Existing Safe Browsing, SmartScreen, and VirusTotal feeds answer “is this site malicious for a human?” — not “is this content safe for an AI agent to ingest?”
CSS and off-screen hidden text
display:none, visibility:hidden, opacity:0, font-size:0. Invisible to a user, read verbatim by an LLM.
Unicode-tag and zero-width encoding
Instructions encoded in Unicode tag characters (U+E0000–E007F) or zero-width spaces. Browsers render nothing; AI contexts ingest everything.
Metadata, comments, and structured data
JSON-LD, Open Graph tags, HTML comments, and schema markup are never shown to a human visitor. They can carry arbitrary instructions for an AI reader.
CyberArmor evaluates external content before it enters AI context, then allows, warns, redacts, sandboxes, blocks, or isolates based on tenant policy — with evidence written on every non-cached decision.
URL Trust Gate runs end-to-end.
The 15-minute local PoC installer brings up the full gate stack on any developer laptop and submits four crafted attack pages — benign, CSS-hidden promptware, zero-width injection, credential-harvest — all producing live verdicts in under 120 ms.
Implemented and Tested
- FastAPI service — end-to-end, health-checked, Prometheus metrics on port 8014
- URL canonicalization, querystring redaction, homoglyph / punycode normalization
- SSRF-guarded safe crawler — no cookies, no credentials, redirect-hop revalidation
- Playwright detonation sandbox on an isolated Docker network (port 8015, internal)
- Heuristic detection: prompt injection, credential harvest, brand impersonation, zero-width stripping
- ML-based detection via detection service — phishing, promptware, DLP, IOC scoring
- Google Safe Browsing v4, Microsoft SmartScreen, and VirusTotal v3 reputation feeds
- Tenant allow/block lists via policy service (GET /policies?tenant_id=…&scope=url-trust-gate)
- Policy decisions: allow, warn, redact, sandbox, block, isolate
- Evidence writes to audit service on every non-cached decision
- Consumer hooks: browser extension, endpoint agent, RASP Python, LangChain SDK, LlamaIndex SDK
Evidence Is Strongest When It Is Attached to Control.
CyberArmor.AI is built around one operating loop: identify the AI action, inspect the risk, enforce the right control, and preserve evidence that explains what happened. Logs alone are not enough. Controls without proof are hard to defend. The platform is designed to bind both together.
Identify
Attribute the AI action to a tenant, user, app, workload, agent, provider, and model where the deployment path provides that context.
Inspect
Evaluate prompt risk, sensitive data, credential leakage, provider posture, and policy-relevant context before the action is trusted.
Control
Apply the approved response: monitor, warn, block, route, limit, or redact in supported browser, endpoint, SDK, extension, and RASP paths.
Connect
Preserve the relationship between actors, agents, tool use, policy decisions, response actions, and downstream evidence records.
Prove
Produce reviewable decision-level evidence for SOC teams, AppSec, legal, compliance, audit, and executive stakeholders.
Pre-Breach Data Protection Example
Redact the secret before it becomes an incident.
Credential leakage into generative AI is a practical enterprise risk. CyberArmor.AI redaction modes are designed to turn policy into an optional response action: remove supported secrets, PII, PCI, NACHA, NPI, or non-public indicators while preserving evidence about the control that ran.
BEFORE
Summarize this log: OPENAI_API_KEY=sk-... and password=hunter22AFTER
Summarize this log: OPENAI_API_KEY=[REDACTED-OPENAI-KEY] and [REDACTED-PASSWORD]Evidence: decision=ALLOW_WITH_REDACTION, policy=redact-secrets, findings=OPENAI_API_KEY/PASSWORD, raw secret preview suppressed.
A Clear Boundary Between What Is Deployable and What Is Being Expanded.
CyberArmor.AI is being built with security-led design partners and controlled pilot deployments. The platform already includes working control, detection, policy, redaction, routing, identity, audit, endpoint, and onboarding paths. Broader enterprise workflows are intentionally marked as pilot-stage while they mature.
Available Today
- URL & Context Trust Gate — 15-minute local PoC with four live attack-page demos (display:none promptware, zero-width injection, credential harvest, benign)
- SSRF-guarded safe crawler and optional Playwright detonation sandbox for JavaScript-heavy content
- External reputation feeds: Safe Browsing v4, Microsoft SmartScreen, and VirusTotal v3 with in-process caching
- Tenant URL allow/block lists with scope-based policy integration
- AI request monitoring and policy decision logging
- Prompt-risk, credential leak, and sensitive-data inspection
- Redaction-mode policy decisions for supported browser, endpoint, SDK, extension, and provider paths
- AI provider routing, provider resolution, credential-handling, and audit event patterns
- Agent identity registration, tenant scoping, delegation metadata, and revocation paths
- Endpoint-assisted AI tool and connection discovery
- Tenant-scoped policy builder, artifacts, and API-key flows
- Audit logs, telemetry, incidents, and evidence capture
In Pilot / Design Partner Phase
- Expanded shadow AI inventory across SaaS and identity sources
- Agent trust verification and delegation-chain workflows
- Proof Vault evidence packaging and external audit exports
- Production SIEM/SOAR integration workflows
- Advanced fail-closed runtime enforcement across more enterprise control points
- Native PQC deployment validation where customer environments support required cryptographic dependencies
- Industry-specific compliance and reporting packs
URL Trust Gate runs end-to-end in 15 minutes on a developer laptop — detection, policy, evidence, and four live attack-page verdicts.
Audit, action graph, compliance, and evidence paths are tied to enforcement decisions, not just reporting copy.
Endpoint agents, browser/IDE/Office extensions, SDKs, RASP hooks, and LangChain/LlamaIndex wrappers support pilot validation.
Built to Map AI Controls to the Language Buyers Already Use.
CyberArmor.AI is designed to help security teams translate AI governance into runtime controls and evidence records that can be reviewed against recognized AI security, risk, and management-system frameworks.
OWASP GenAI Security
Prompt injection, sensitive information disclosure, excessive agency, supply chain risk, and model theft.
NIST AI RMF
Govern, map, measure, and manage AI risk through an operational risk-management model.
ISO/IEC 42001
AI management system practices for organizations building, using, or governing AI systems.
EU AI Act
Documentation, logging, transparency, human oversight, robustness, accuracy, and cybersecurity expectations.
Built for the Realities of Enterprise AI Risk.
Every capability maps to the integrated runtime loop: identify AI activity, inspect risk, decide policy, enforce the approved response, and prove what happened.
URL & Context Trust Gate
Pre-ingestion safety check for URLs and external content before it reaches any human, browser, endpoint agent, or AI agent. Detects phishing, hidden prompt injection, promptware, and IOCs — including CSS-hidden and zero-width-encoded payloads. Backed by Safe Browsing v4, Microsoft SmartScreen, and VirusTotal reputation feeds. 15-minute local PoC available.
Prompt Injection Defense
Detect prompt injection, jailbreak attempts, adversarial normalization patterns, and suspicious prompt behavior targeting AI applications and workflows.
Shadow AI Discovery
Surface unreviewed AI tools, model calls, browser usage, endpoint activity, APIs, and service connections through supported signals.
AI Agent Trust & Control
Register AI agents, scope allowed tools, track delegation context, and preserve evidence around autonomous workflows as trust controls mature.
Sensitive Data Redaction
Inspect AI-bound data for credentials, secrets, PII, PCI, bank-routing data, healthcare identifiers, and non-public indicators so supported paths can redact, warn, log, or block by policy.
Identity-Aware Policy Engine
Apply contextual controls to humans, services, workloads, and AI agents with tenant-scoped policy evaluation and decision records.
Provider Routing & Control
Resolve approved AI providers, route OpenAI-compatible and Anthropic-style requests, handle provider credentials, track cost signals, and emit audit events.
Action Graph & Evidence
Capture trace IDs, policy decisions, actor context, delegation chains, data classifications, signatures, and chain hashes for reviewable AI activity records.
Runtime Response
Connect detection and policy decisions to response actions such as block, redact, route, notify, limit, revoke, or hand off into SOC workflows.
This Is What Different Looks Like.
The AI security category is getting crowded with point products and retrofitted tooling. CyberArmor.AI is aimed at the harder problem: making AI activity controllable, attributable, and provable across the places where enterprise AI actually runs.
Governance That Becomes Runtime Control
Most governance products stop at visibility, questionnaires, or policy documentation. CyberArmor.AI ties governance to runtime decisioning so policy can result in monitor, warn, block, route, limit, or redact actions where supported.
Protection-Backed Evidence
Evidence is valuable because it is bound to control. CyberArmor.AI records who or what acted, which policy applied, what response ran, what data classification was involved, and why the decision happened.
Cross-Layer Causality by Design
CyberArmor.AI brings together detection, policy, AI routing, endpoint signals, browser and IDE workflows, agent identity, secrets, response, audit, and compliance evidence so teams can reconstruct the chain.
Honest Platform Boundaries
Enterprise buyers trust specific claims. CyberArmor.AI separates pilot-ready capabilities from roadmap expansion, because a security runtime has to be credible before it can be trusted.
Pre-Ingestion Web Content Trust for AI Agents
Existing URL filters answer 'is this site safe for a human?' CyberArmor.AI also answers 'is this content safe for an AI agent to ingest?' The URL Trust Gate evaluates external destinations for promptware, hidden prompt injection, phishing, IOCs, and credential harvesting before content ever reaches AI context — with policy-based enforcement and evidence written to audit.
Real Threats. Real Buyers. Real Answers.
CyberArmor.AI is built around the security challenges enterprise teams are already facing: AI data leakage, prompt misuse, agent identity, provider sprawl, and the lack of proof when AI activity crosses a boundary.
Shadow AI Discovery & Governance
The Problem
Employees and vendors are connecting to AI tools, APIs, and models without security review. You have limited inventory, uneven controls, and little evidence when usage crosses policy.
The Solution
CyberArmor.AI uses supported endpoint, browser, API, and integration signals to surface AI usage, connect it to policy, and preserve evidence as coverage expands with each deployment path.
From AI Activity to Controlled Evidence.
CyberArmor.AI follows a practical runtime model: attribute the AI action, inspect risk, enforce the policy, and preserve evidence. That is the loop security teams need to make AI governance operational.
Not Just Logs.
Control You Can Prove.
When something goes wrong with AI — and it will — security teams need more than alerts. They need a structured, reviewable record of the control decision: which user or agent, which provider or model, what data was involved, which policy applied, and what response ran.
CyberArmor.AI captures evidence as part of the runtime control loop. That makes investigations stronger because the proof is connected to the action: blocked, redacted, warned, routed, limited, or allowed.
Zero-width prompt injection detected in external page — AI agent fetch blocked before content ingestion
Credential removed before AI-bound browser prompt submission
Request routed to approved model path with credential handling
Built by a Security Practitioner for Teams That Need Control and Proof.
CyberArmor.AI is founder-led by Patrick Kelly and built from the operating reality of enterprise security: policy has to become enforcement, sensitive data has to be protected before exposure, incidents need evidence, and AI adoption cannot wait for a perfect governance program.
The company is intentionally transparent about product maturity, design-partner work, and where the platform is strongest today. That posture is part of the product.
Read the company storySecurity practitioner
Built from application, data, cloud, endpoint, identity, and AI security operating problems.
Hands-on builder
Rooted in working controls, tests, demos, runbooks, and deployment paths instead of slideware.
Enterprise lens
Designed for regulated environments, uneven ownership, legacy systems, and real security-team workflows.
Official Brand and Domains
CyberArmor.AI is the public brand and product site operated by CyberArmor AI, Inc. Official CyberArmor.AI web properties are served from cyberarmor.ai and its subdomains, including app.cyberarmor.ai, admin.cyberarmor.ai, docs.cyberarmor.ai, and support.cyberarmor.ai. CyberArmor.AI is not affiliated with similarly named third-party domains, services, or social profiles unless they are explicitly linked from one of those official properties.
Built for pilots. Honest about the boundary.
Security buyers deserve to know exactly what is working end-to-end, what requires configuration, and what is on the roadmap. Here is where CyberArmor stands today.
Pilot-ready
URL Trust Gate runs end-to-end. 15-minute local PoC available. Three optional reputation feeds configurable via environment variables.
Deployable today
The broader platform — control plane, policy, detection, response, secrets, and endpoint agent — is deployable for controlled pilots, internal deployments, and operator-led staging environments.
Still maturing
Some customer-facing SaaS workflows, self-service onboarding, and MFA enforcement are still being refined and are marked as such in the capability status table.