Prompt inspection
Tuned, quantized local AI models identify sensitive content. The current PoC validates text inspection and basic hard blocking.
VALIDATED IN LOCAL POC SUPERVISORY INTELLIGENCE / RISKORAI
Let AI work.
Keep data yours.
A local-first prompt security gateway in development. Designed to inspect AI requests, block sensitive content, and give security leaders a clearer view of enterprise AI use.
DARK MODE / YOUR CALL
LOCAL INTELLIGENCE
HUMAN OVERSIGHT
HUMAN OVERSIGHT
THE NEXT FRONTIER
SECURE BY INTENT.
SECURE BY INTENT.
PROMPT SECURITY / AT THE ENTERPRISE EDGESTATUS: PROOF OF CONCEPT
INSPECT / LOCAL AI MODELSENFORCE / HARD-BLOCK POCDEPLOYMENT / EDGE APPLIANCE PLANNED
THE EXPOSURE
Your next data leak
could start with a prompt.
Employees use generative AI to move faster. Customer records, internal reports, and proprietary code can travel with them. RiskorAI focuses on inspecting this conversational boundary.
Context-aware protection
Planned placeholder substitution will mask sensitive data, with hard blocking when redaction would damage the prompt’s context.
PLANNED / PHASE 2Governance & audit
Base administrative logging is established. Centralized policies, reporting, and a tamper-evident audit vault form the next development scope.
BASE LOGGING / POCSEE THE DECISION FLOW
A boundary before
the send button.
Switch between two predefined examples to understand the hard-block concept. This is an interactive illustration, not a running AI security engine.
RISKORAI / PROMPT INSPECTIONCONCEPT PREVIEW · NO NETWORK REQUESTS
EXAMPLE PROMPT
Summarize the customer record for Maya. National ID: 3273014405900001.
ILLUSTRATED OUTCOME
Request blocked
This example contains a personal identifier. In the PoC hard-block flow, a flagged request is rejected before forwarding.
Synthetic example data. No prompt is submitted, analyzed, or stored by this page.
BUILT AROUND YOUR ENVIRONMENT
Local intelligence.
Enterprise oversight.
Designed for CISOs, risk leaders, and data protection teams in organizations with managed or corporate-owned devices—especially banking, financial services, telecom, and B2B SaaS.
Validate the engine
Local Windows prototype, lightweight model execution, basic hard blocking, and foundational administrative logging.
Test the perimeter
Migrate to a dedicated TLS forward-proxy appliance and validate interception, latency, and operational behavior in managed environments.
Extend governance
Add contextual sanitization, semantic anomaly detection, tamper-evident logs, and centralized security policy management.
Does RiskorAI currently run as an enterprise network appliance?
No. The current implementation is a local Windows proof of concept. A network-edge appliance with SSL/TLS forward proxy is a planned migration. Deployment will require approved enterprise device and network configuration.
Does prompt inspection rely on an external AI API?
The PoC uses local, optimized AI models for inspection. The proposed appliance architecture also keeps security parsing local. Permitted prompts may still be sent to the organization’s chosen external AI service.
Is the 95% leak-prevention figure a validated result?
No. The brief states a target of up to 95% prevention of accidental leaks. It does not supply a measured benchmark or test methodology, so this page makes no verified detection or prevention claim.
What is the proposed commercial model?
The planned model combines an enterprise appliance license with a subscription based on prompt filtering volume or throughput. Pricing, service levels, integrations, and production availability are not finalized.
Does RiskorAI guarantee regulatory compliance?
No. RiskorAI aims to support data governance and evidence collection. Compliance depends on an organization’s full technical, legal, and operational controls, and requires independent assessment.
START A CONVERSATION
Shape a safer AI perimeter.
Exploring AI governance for your organization? Talk to our founding team about your environment and potential pilot requirements.