AI TRISM and How Verta Supports It | Verta.ai

Enterprises are finding increasingly innovative ways to leverage AI for intelligent customer experiences. But technology research from Gartner cautions that companies must become rigorous in applying what it calls AI TRiSM to achieve success with their AI initiatives.

Gartner defines AI trust, risk and security management (AI TRiSM) as a “framework that supports AI model governance, trustworthiness, fairness, reliability, robustness, efficacy and privacy” — and named it a top strategic technology trend for 2023. 

According to the research, organizations that put AI TRiSM into practice can anticipate significantly better business outcomes from their AI projects. On the other hand, organizations that don’t take steps to manage AI risk are much more likely to see models not performing as expected, security failures, financial losses, and damage to reputation.

Verta’s Operational AI platform and Enterprise Model Management system include comprehensive capabilities that support AI trust, risk and security management:

Trust

Explainability - the ability to understand how a model arrived at an outcome - is a key component of ensuring trust in AI. 

• Verta supports explainability by providing a central enterprise model management system where data scientists can publish all model metadata, documentation, and artifacts. 
• With Verta, organizations also get visibility to the data supply chain, since Verta tracks model lineage back to training data used in experimentation. 

Risk

Enterprises manage AI risks by applying rigorous governance to models throughout the ML lifecycle. 

• Verta lets you record and manage a model across its lifecycle, from development and staging to production and archive. Companies can set up explainability and bias checks as part of the release process to ensure compliance with Ethical AI standards. 
• Verta enables model validation to ensure that the organization’s models are performing as designed, identifies edge cases that require manual processing or further review, and monitors the overall service health of models through operational metrics such as response time, latency, error rate and throughput.
• Governance and risk teams use Verta to monitor model I/O and performance and administrate governance rules. 
• Verta automatically monitors data quality and drift and model performance metrics like accuracy, precision, recall, and F1 score. Alerts can be set in the event of performance degradation or drift.

Security management

Data protection and overall IT security of the ML process are essential for supporting AI TRiSM. Verta's security management capabilities include:

• Customization of granular access controls for the entire ML lifecycle, providing users the right access to the right information
• Easy integration with your existing security policies and identity management system
• Standardized safe release practices with release checklists and CI/CD automations
• Model scanning for vulnerabilities as part of the deployment process.
• Detailed audit logs available for compliance
• Isolated workspaces between environments, whether between teams or between development and production models

With proposed AI regulations like the ADPPA and the AI Bill of Rights creating new risks for organizations that rely on AI/ML, now is the time to ensure that you are putting the necessary technical capabilities in place to support AI TRiSM in your ML operations.

Contact Verta to discuss how your organization can leverage an Operational AI platform and Enterprise Model Management system to meet the challenges of AI trust, risk and security management.