Artificial Intelligence Community

 View Only

  • 1.  AI in Compliance: Governance, Validation & Risk

    Posted 19 days ago

    As institutions explore how to utilize AI for financial anomaly detection, conflict of interest reviews, and the development of custom systems for research units, how are you structuring governance, validating outputs, and managing liability and compliance risks to ensure these tools enhance oversight without introducing new audit or legal exposure?



    ------------------------------
    Wendy Ying Yu
    Cost Analysis Manager
    University of Maryland, College Park
    College Park, MD
    (301) 405-9283
    ------------------------------


  • 2.  RE: AI in Compliance: Governance, Validation & Risk

    Posted 18 days ago

    We used AI to detect some spending fraud, but our main use case is still document reading (chatbot for policies, information extraction etc). Would also be interested in AI governance discussions.



    ------------------------------
    Youyou Cheng
    Deputy Director, SPFA
    Yale University
    25 Science Park
    New Haven, CT 06511
    203-785-4370
    youyou.cheng@...
    ------------------------------

    Original Message


  • 3.  RE: AI in Compliance: Governance, Validation & Risk

    Posted 13 days ago

    Thank you for the input, Youyou. During our group discussion, some members shared their instituion are using AI for COI reviews or have developed customized AI system for vaccin research units. One member was particularly interested in how AI can be used to analyze fiannncial date transaction anomaliies which seems to align with the use case you described. Could you share a bit more about the AI system used to detect spend fraud?

    In the context of Ai Governance,while enterprise paid AI is a sigiicantly safer and more efficient choice for organizations, it is still subject to risks. Even with robust protections, systems remain vulerable to threats like prompt injection, human error. requiring ongoing monitoring and oversight. 



    ------------------------------
    Wendy Ying Yu
    Manager, Financial Compliance
    University of Maryland, College Park
    College Park, MD
    (301) 405-9283
    ------------------------------

    Original Message


  • 4.  RE: AI in Compliance: Governance, Validation & Risk

    Posted 12 days ago

    My office is utilizing a closed-instance LLM to conduct AI reviews of documents as a "second look" for our review of complex comparisons of documents. The reason we chose this task is because it is time consuming for us to complete, susceptible to human error, and because we have clear SOPs written on them that we were able to easily translate into a prompt. 

    Youyou (and others who have implemented AI as part of their processes), what has been your experience with the accuracy of the work done by AI? 

    After much refining, we've gotten to a place where the outputs we're getting for our particular project are consistently helpful and largely accurate, but it took months of work and for the initial setup. It actually took more time to complete the task initially because of the significant attention we had to pay to the AI accuracy, and the results are still something we approach with caution.



    ------------------------------
    Leah Gessel
    Conflict of Interest and Conflict of Commitment Administrator
    University of Oregon
    Eugene, OR
    (541) 346-1408
    ------------------------------

    Original Message


Related Content