Dernière mise à jour :
AI model risk is the risk that an AI system produces incorrect, biased, or unstable outputs that lead to material financial, operational, or reputational harm. As AI models are deployed across underwriting, credit decisioning, trading, and portfolio management, the risks embedded in those models become systemic risks for the institutions that use them—and for the markets they operate in. Earthian's Technology Tenet-0 is the first AI model purpose-built to analyze and frame the risks of other AI systems. Risk management connects these capabilities to governance, limits, mitigation actions, and continuous monitoring through Earthian Hub.
Frequently Asked Questions
AI model risk is the risk that an AI system produces incorrect, biased, or unstable outputs that cause material harm. It includes model drift, adversarial vulnerabilities, training data contamination, goal misspecification, hallucination, and governance failures. AI model risk is structurally different from traditional model risk—requiring inference-driven assessment rather than historical backtesting.
Technology Tenet-0 infers AI model risk from technology architecture disclosures, patent filings, SEC documents, and operational signals—without requiring direct model access. It identifies structural vulnerabilities, drift conditions, adversarial exposure, and governance gaps through inference rather than direct inspection.
Traditional MRM frameworks were designed for statistical models and rely on backtesting, annual validation cycles, and pass/fail conclusions. AI systems operate on fundamentally different principles—they drift, hallucinate, and fail under adversarial inputs in ways that historical backtesting cannot detect. Continuous, inference-driven AI model risk management is required.
Systemic AI model risk is the risk that AI failures in one institution propagate across interconnected financial systems. When multiple institutions use similar AI models—or AI models trained on the same data—correlated failures can create systemic market events. Technology Tenet-0 and Earthian Hub enable cross-system propagation analysis that traditional siloed MRM frameworks cannot perform.
Financial institutions face AI model risk across credit decisioning (biased or drifted models producing adverse selection), trading (AI models amplifying volatility through correlated behavior), underwriting (generative AI models hallucinating risk assessments), and compliance (AI governance failures attracting regulatory action). Technology Tenet-0 addresses each of these exposure dimensions.