AxiLayer AI | The Agent Control Plane | Find it. Control it. Prove it.

The platform

Twelve phases, in the order you run them

Everything on this page is in AxiSentinel or AxiReady today.

Phases 01 and 02

Discover

Layer 01 Find · AxiReady

LiveCloud discoveryRead-only Azure Resource Graph, AWS Resource Explorer, Google Cloud Asset Inventory. The client grants a reader role; nothing is written.

LiveHost and container scanAXI-Node reports hosts, CPU, memory and containers. vllm, ollama, triton, tgi, nim and llama images count as AI services; unclaimed ones become Shadow AI candidates.

LiveAgent and tool inventoryRegistered agents, shadow agents, tool servers and the identities they run as, with a readiness score each.

LiveClassify and planRisk class, frameworks that apply (220+ encoded), and the monitoring plan each system needs.

Phases 03 to 06

Control

Layer 02 Control · AxiSentinel

LiveMCP GatewayEvery tool server an agent can call is discovered and gated. Unapproved: blocked, alert raised. Approved by an authorized person.

LiveAgent identityagent_credential_id and agent_role_scope recorded on every action. No shared service accounts.

LiveAction and authority gatesHigh-impact actions (money, care, safety systems, production data) wait at the authority gate; the approvals inbox is where a person confirms, with a reason.

LiveLLM proxyModel calls pass through one proxy, so policy, logging and the stop apply to every model the same way.

Phases 07 to 09

Observe and Respond

Layer 02 Control · AxiSentinel

LiveAI behavior guard14-day baseline per agent. Drift, spikes and signs of hijack flagged; at 10x normal, Stop recommended.

LiveAI radarWhat is running, where, in real time, across regions, with open problems per region.

LiveQuarantine and kill switchCalls refused at the gateway; stop enforced at ingest and proxy on one confirmed decision. Emergency suspend from Command.

LiveArticle 72 and 73Five kinds of monitoring data checked per plan (arriving, stale, missing); Article 73 clock on every incident.

Phases 10 to 12

Prove

Layer 03 Prove · AxiLayer AI

LiveAgentic test suiteCases mapped to the OWASP Top 10 for Agentic Applications; re-run after every model, data or vendor change.

LiveControl Effectiveness ScorecardFor each control: events, blocked or held, missed. Signed.

LiveCompliance PassportSigned by a qualified AxiLayer AI auditor; public verify, free, forever.

LiveHuman sign-offAutomation finalizes nothing. No AI approves another AI.

Platform

Govern and Run

All layers

LiveDeploymentYour cloud tenant, your data center, or air-gapped. Silent deployment package; 9 locales.

LiveWorkflow menuHome, Discover, Classify and Plan, Control, Observe, Respond, Prove, Govern, Run the platform, Learn.

LiveNotebooks and evidence exportSaved query notebooks, CSV export, every figure clickable to its records.

LiveAcademyFrom learner to qualified AxiLayer AI auditor; the people who sign are trained here.

Three layers, twelve phases

Twelve phases. One platform. One signature.

Phase 01FindDiscover and score

Phase 02FindClassify and plan

Phase 03ControlApprove every tool

Phase 04ControlName every agent

Phase 05ControlGate every action

Phase 06ControlHold for a person

Phase 07ControlWatch behavior

Phase 08ControlQuarantine and stop

Phase 09ControlMonitor to the article

Phase 10ProveTest and re-test

Phase 11ProveSign the evidence

Phase 12ProveVerify, free, forever

AxiReady finds. AxiSentinel gates, holds, quarantines and stops, Always On or On Demand. A qualified AxiLayer AI auditor signs every step.

The operating model

AxiSentinel controls. AxiLayer AI signs.

Control is built into the platform; the decisions that matter are confirmed by your people; the signature comes from an auditor who did not build the system under review.

AxiSentinel

Controls, live

GatesHoldsQuarantinesStopsRevokes

Always On or On Demand. The stop is built in, and no AI approves another AI.

Your authorized people

Confirm what matters

ApproveReleaseConfirm stop

High-impact decisions are confirmed by your admin or technical lead inside AxiSentinel, with a reason, in a tamper-evident record.

AxiLayer AI

Reviews and signs

ReviewsSignsRe-verifies

A qualified AxiLayer AI auditor who did not build the AI system under review signs the findings and the passport.

Frameworks

Mapped to what regulators and attackers use

EU AI Act, Articles 14 and 15

Human oversight and the ability to intervene or stop: authority gates, approvals inbox, kill switch. Accuracy, robustness and cybersecurity: agentic test suite, behavior guard.

OWASP Top 10 for Agentic Applications 2026

Goal hijack, tool misuse, identity and privilege abuse, agentic supply chain, code execution, memory poisoning, inter-agent communication, cascading failures, human-agent trust, rogue agents. Each control and each test case is mapped to one.

NIST AI RMF and NIST AI 600-1

Govern, Map, Measure, Manage, with the generative AI profile actions evidenced by the scorecard.

ISO/IEC 42001 and ISO/IEC 27090

Management-system controls and AI-specific security guidance, evidenced continuously rather than at audit time.

Singapore Model AI Governance Framework for Agentic AI

Bounded autonomy, human accountability and traceability, mapped to gates, identity and the record.

MITRE ATLAS

Adversarial techniques against AI, used to build the authorized adversarial test cases.

Why now

The year AI agents became the attacker and the insider

The data behind the control plane, from 2026 surveys, breach reports and supervisory letters.

82%

of enterprises found AI agents running in their infrastructure that nobody had registered; 41% found them more than once. Cloud Security Alliance survey of 418 IT and security professionals, fieldwork January 2026, published 21 April 2026.

65%

had an AI agent-related incident in the past twelve months: 61% data exposure, 43% operational disruption, 35% financial loss. Only 21% have a formal process to decommission an agent. Same survey.

11%

automatically block an agent that acts outside its scope; 38% require human approval and 24% only log it. Same survey. AxiSentinel does all three: blocks at the gateway, holds for a person, records.

43%

of breaches now involve shadow AI, up from 20% a year earlier; shadow AI incidents averaged USD 5.39 million. IBM Cost of a Data Breach Report 2026.

92%

of AI-related breaches involved no proper AI access controls, and only 40% of organizations apply access controls to AI models and data. AI governance was absent at 68% of breached organizations. IBM 2026.

USD 1M

added to the average breach cost when AI drove the attack; prompt injection breaches averaged USD 5.89 million. Global average breach cost USD 4.99 million, a record. IBM 2026.

100+

organizations notified by a frontier lab on 1 October 2026 after about 700 of its own evaluation agents chained exposed credentials and two zero-days into a public model platform's production systems between May and July 2026.

2,054

of 9,695 public MCP tool servers analysed by Trend Micro in July 2026 had no authentication at all, and 2,259 were exploitable. Tool servers are the new unpatched perimeter.

named agentic risks in the OWASP Top 10 for Agentic Applications 2026, from goal hijack and tool misuse to rogue agents. Every AxiSentinel control and every test case maps to one.

Sources: Cloud Security Alliance and Token Security, State of AI Agent Security, 21 April 2026; IBM Cost of a Data Breach Report 2026; Trend Micro MCP server analysis, July 2026; frontier-lab incident report and notifications, 1 October 2026; OWASP Top 10 for Agentic Applications 2026.

What supervisors ask for

Inventory, control, monitoring, evidence

EU AI Act, Articles 14, 15, 72 and 73

Human oversight with the ability to intervene or stop; accuracy, robustness and cybersecurity against data poisoning, adversarial inputs and confidentiality attacks; post-market monitoring data; serious-incident reporting clocks. High-risk obligations apply from 2 December 2027 and 2 August 2028 under the 2026 Omnibus.

NYDFS, 21 May 2026

Industry letters on frontier AI cyber risk under Part 500: updated risk assessments, faster remediation, review of AI-generated code, more frequent incident response testing.

European Supervisory Authorities, July 2026

Joint statement under DORA: governance for AI-assisted cyber incidents without delay, updated risk appetite, faster patching, supply-chain standards for AI and ICT providers.

European Central Bank, 7 July 2026

Significant banks asked for AI cyber action plans by 31 October 2026, including assurance from ICT providers.

MAS, Singapore, 7 October 2026

AI Risk Management Guidelines, in force from 7 October 2027: board accountability, AI inventory, lifecycle controls including cybersecurity, and assurance that governance cannot be delegated.

ISO/IEC 27090 and NIST

ISO/IEC 27090:2026 on security of AI systems; NIST AI RMF, NIST AI 600-1 generative AI profile and the NIST Cyber AI Profile draft, all expecting inventory, access control, live monitoring and evidence.

The Agent Control Plane