AI Research Summary
Before Human ReviewBased on abstract · Full text not reviewed
This is an abstract-based summary, and the full text was not reviewed. It argues that as financial institutions increasingly use ML and generative AI, complex risks arise that cannot be addressed by existing information security controls and traditional model risk management (MRM) alone. The authors propose AI-SMRG, an integrated framework that considers threats such as adversarial perturbations, data poisoning, model extraction, membership inference, and prompt injection together with model drift, overfitting, and insufficient validation.
Key Points from the Abstract
- It argues that the use of AI in finance creates a new attack surface in which security and model risk are combined.
- AI-SMRG integrates cybersecurity engineering and MRM into a single, continuous monitoring system.
- It argues that operating traditional security controls and MRM in parallel is not sufficient.
Relevance to AI Law and Policy
It is directly relevant to the integrated regulatory design of financial AI governance, cybersecurity, and model risk management.
Limitations to Consider
- Method, evaluation, and empirical results cannot be confirmed from the abstract alone.
- The effectiveness of the proposed framework is not demonstrated in the abstract.
- Limitations are not specified in the abstract, so no additional limitations can be identified.
Please compare with the original text before citing or using it in your assessment.
Abstract Preview
Financial institutions increasingly depend on machine learning and generative artificial intelligence to underwrite cred…Read more in the abstract and original text Research Topic
Original text and source
AI SECURITY AND MODEL RISK GOVERNANCE
The Korean title on this page was translated by AI.- Database
- Crossref