Asenion AI Management System

The Asenion AI Management System is an award-winning Governance, Risk and Compliance + Assurance platform providing AI Trust, Risk and Security Management (AI TRiSM) solution. We perform proactive assurance and risk reduction for predictive, generative and agentic AI systems by combining an AI Risk Navigator and AI Compliance Monitor with a collection of Assurance Packs to provide posture management, AI system adversarial testing, AI validation, third-party risk management and monitoring against compliance by defining policies and controls.

Asenion AI TRiSM Systems Diagram

Asenion Platform Overview

The Asenion AI Management System is a web-based interface designed to streamline AI trust, security, and compliance efforts. It acts as a central hub for users to assess their AI system’s compliance to selected policies and determine their inherent and residual risk status by looking at likelihood and impact of each type of risk as defined by these policies. Policies are recommended based on sector, jurisdiction, methodology, use case and where the AI system is in its project lifecycle. Moreover, policies are mapped to qualitative and quantitative controls which are used to assess societal, organizational, system, model, dataset, generated content and individual’s risks. The platform facilitates collaboration among multiple stakeholders by enabling them to share projects, thus streamlining assessments. Additionally, it allow users to create and disseminate reports to stakeholders for feedback and approval.

Key components of the Asenion Platform include:

  • Use Case: A use case represents an AI system that is predicitive, generative and agentic. It can be developed internally or externally by a third-party. Learn more

  • Policy Pack: A policy pack represents a qualitative assessment based on internal and external policies, frameworks, regulations and standards Learn more

  • Policy Library: A collection of policy poacks. Learn more

  • Assurance Pack: An assurance pack represents a policy-aware qunatitative test and test agent (a microservice that can be deployed in standalone envrionments) based on internal and external frameworks, benchmarks, regulations and standards Learn more

  • Assurance Library: A collection of assurance poacks. Learn more

  • Control Bundle: A control bundle represents a category of controls within a policy. ‍ Learn more

  • Control: A control represents a qualitative or quantitative metric within a control bundle within a policy. Controls can be proactive, detective or preventative. Learn more

  • Role-based Access: A admin user role in the system has additional priviledges. Learn more

  • PersonaGen: PersonaGen is a gen AI tool used to create persona with fake memories - a reality-rooted dataset for testing LLM and Agentic AI applications. Learn more

  • PolicyGen: PolicylGen is a gen AI tool used to create new policy assessments. Learn more

  • ControlGen: ControlGen is a gen AI tool used to create new controls. Learn more

  • AnswerGen: AnwserGen is a gen AI tool used to generate answers for controls in Policy Packs. Learn more

  • TestGen: TestGen is a gen AI tool used to create new test cases for LLM and LLM+RAG applications. Learn more

  • Lifecycle Management: The Asenion Lifecycle Management features allows organization to configure and customize their own AI model lifecycle management stages with automated approval thresholds. Learn more

  • Ongoing Monitoring: The Asenion ongoing monitoring feature allow users to monitor individual controls against their policy alignment and risk thresholds over time Learn more

Asenion Assurance Framework Overview

Asenion Assurance is both a framework and a collection of test agents designed to automate compliance testing.

Often regarded as a secondary, ad-hoc task, compliance testing typically occurs when compliance reports are necessary. This task can be both time-consuming and tedious. Asenion’s aim is to expedite and standardize this process, making it more efficient.

Within the Asenion platform, an Assurance Pack consists of various control bundles, each composed of a set of quantitative controls necessitate the execution of model tests to satisfy the control’s criteria. For instance, the accuracy_score is a performance control utilized to evaluate the precision of a predictive model.

Asenion Assurace framework supports the integration of various test agents into a uniform testing system. An Assurance Pack outlines the quantitative controls necessary for the project. Leveraging this Assurance Pack, the Asenion Assurance framework activates the appropriate test agents. These agents conduct the necessary tests, compile the results, and forward them to the Asenion Platform for further analysis.

See the Test Frameworks section for tests provided by the Asenion agents.

Asenion Platform and Asenion Integration

Asenion AI Governance Platform and Asenion Systems Diagram

The integration process between the Asenion Platform and Asenion involves a sequence of steps designed to ensure seamless communication and data exchange. Here’s a streamlined description of this process:

  • Policy Specification: Initially, a policy delineates a comprehensive list of controls or tests that need to be executed.
  • Configuration Download: Through the project-info API, Asenion can obtain a test configuration file specific to a project from the platform.
  • Asenion Test Runner Setup: This configuration file is then installed within the Asenion Test Runner, preparing it for the testing phase.
  • Test Execution: With the configuration in place, the Asenion Test Runner activates the relevant test agents. These agents proceed to execute the specified tests as per the project’s requirements.
  • Result Compilation and Upload: Upon completion of the tests, the test agents compile the outcomes into a JSON format. These results are then directly uploaded back to the corresponding project on the Asenion Platform via the test-results API, completing the integration loop.

This integration framework ensures that tests are conducted efficiently and the results are seamlessly fed back into the Asenion Platform, enhancing the overall compliance and risk assessment process for AI and predictive model systems.

Learn more