In this study we examine how ChatGPT makes asset allocation decisions based on simulated data. We feed 100,000 hypothetical investor profiles into ChatGPT and ask it to recommend an asset allocation between a diversified equity fund and a diversified bond fund for each investor (client). We include a rich set of variables for each client such as their age, income, race, risk aversion, financial knowledge, and health. We instruct ChatGPT to act as a financial advisor. However, we do not require it to prioritize any set of variables over others. Our evidence indicates that ChatGPT recommends a higher equity allocation for those with less risk aversion and for those who are more optimistic, confident, and financially knowledgeable. Our evidence does not indicate that ChatGPT uses client demographic data such as age, race, gender, and marital status in forming its recommendations. Our results suggest that ChatGPT may allow investor biases (such as overconfidence) to influence its asset allocation recommendations. ChatGPT may also inadvertently hurt the financial success of clients with less financial knowledge by steering them away from equities. Since it ignores client age, ChatGPT may also recommend allocations that are too conservative for some of the young investors and too aggressive for some of the old investors. This study contributes to the literature on the applications of Artificial Intelligence (AI) to portfolio allocation and household financial management decisions.

Federated learning (FL) is a revolutionary machine learning technology that protects data ownership while training unified AI models. By enabling multiple organisations to train machine learning models collaboratively by exchanging model updates instead of raw data, federated learning systems have great potential in areas where raw data are sensitive and cannot be easily shared, such as in financial technology (Fintech). Federated learning also emerges as a novel approach in the domain of anti-fraud systems to identify and combat economic crimes, such as fraud and money laundering, without sacrificing the security of sensitive financial information. This paper discusses recent developments using federated learning for Fintech and highlights its application in combatting fraud in Taiwan. Federated learning has successfully optimised fraud detection models across multiple financial institutions, as evidenced by key projects like the "Eagle Eye Fraud Detection Alliance Platform". Such initiatives prove that FL can significantly improve early fraud detection across institutions while ensuring data privacy through joint training of AI models. It also outlines a brief overview of security issues, Vision for Federated Learning, and the major challenges seen in widespread adoption, such as issues in model inversion attacks, data heterogeneity, and the robust encryption methods that can make it work. However, these problems do not outweigh the advantages of using federated learning to improve Fintech anti-fraud mechanisms. This paper then concludes with a discussion on possible future work and the usage of FL to also improve financial crime detection, presenting novel opportunities for institution-wise cooperation and a more effective anti-fraud scheme.

The emergence of cryptocurrencies and artificial intelligence (AI) has had a major effect on worldwide financial crime, allowing for increasingly complex fraud schemes. Criminal groups increasingly leverage these technologies, bundling them with sophisticated business models such as phishing-as-a-service and ransomware-as-a-service. The rapid growth of AI and cryptocurrencies has generated a sharp increase in organized financial crime targeting vulnerable people across borders, according to a recent INTERPOL assessment. Organized crime syndicates are using AI to create convincing deepfakes to dupe victims into false investment or romance scams and cryptocurrencies to transfer ill-gotten gains anonymously. One example is “romance baiting” fraud, in which trafficked people, often under duress to commit crimes, use AI-generated faces to dupe victims into financial scams. This hybrid romance and investment fraud scam has become widespread in Southeast Asia, Africa, and Latin America. Criminal networks repurpose government intelligence tools to target these narratives, making it difficult for authorities to detect and intervene. Finally, this paper examines the implications of  new fraud techniques for law enforcement and the urgent need for greater international cooperation and regulatory frameworks. It underlines the promotion of data exchange, capacity-building, and public–private partnerships as crucial mechanisms for tackling AI-powered financial fraud. This paper further highlights the need for consumer awareness for reporting mechanisms to tackle the growing risks of cryptocurrency and AI-enabled scams.

Modelling and forecasting cryptocurrency volatility is essential due to the inherently volatile and speculative nature of digital asset markets. Accurate volatility predictions enable traders and investors to make informed decisions, optimize portfolio strategies, and mitigate risks in a highly uncertain environment. This study examines the volatility dynamics of large-cap and mid-cap cryptocurrencies through high-frequency data analysis. Cryptocurrencies exhibit unique market behaviours characterised by complex short-, medium-, and long-term volatility patterns, which require sophisticated modelling techniques for accurate forecasting. Among the methods explored, the Heterogeneous Autoregressive (HAR) model stands out for its ability to effectively capture multi-scale dependencies, making it particularly suitable for modelling the complex volatility trends observed in these digital assets. By assessing both in-sample and out-of-sample performance, this study points out the importance of employing multi-scale approaches to improve predictive accuracy. The findings have significant implications for risk management and trading strategies, as accurate volatility forecasting is crucial in highly volatile cryptocurrency markets. The HAR model’s capacity to integrate multiple time horizons allows for a more comprehensive understanding of market dynamics, providing practical insights for financial decision-making. This research advances the broader understanding of cryptocurrency volatility and provides a foundation for future studies to explore understudied modelling approaches to the growing complexities of digital asset markets.

Introduction:
Artificial Intelligence (AI) is transforming financial reporting and auditing by automating complex processes, improving accuracy, and enhancing decision-making capabilities. Current advancements have demonstrated significant potential to streamline operations, increase transparency, and reduce human error. However, this transformative shift is accompanied by risks, such as algorithmic bias, cybersecurity threats, and regulatory challenges. This study examines the current applications of AI in financial reporting and auditing, explores potential future uses, and identifies associated opportunities and risks.

Methods:
This research employs a qualitative analysis of industry case studies to explore AI applications in financial reporting and auditing. Data were collected from published reports, industry insights, and case studies detailing the implementation of AI-driven tools by auditing firms and corporate finance departments. This study categorizes existing AI technologies and projects their future applications based on current trends and expert forecasts.

Results:
Key AI technologies in use today include natural language processing for automated report generation, machine learning for fraud detection, robotic process automation for data reconciliation, and predictive analytics for forecasting financial trends. Potential future applications encompass real-time auditing powered by AI-enhanced blockchain systems, advanced anomaly detection using deep learning algorithms, and AI tools for assessing compliance with increasingly complex regulatory requirements. These developments promise significant efficiency gains, but also highlight risks such as ethical concerns over algorithm transparency and challenges in safeguarding sensitive financial data.

Conclusions:
AI is revolutionizing financial reporting and auditing, offering unprecedented opportunities to improve efficiency and accuracy. However, its implementation must be accompanied by careful consideration of ethical, regulatory, and security concerns. Establishing robust frameworks for AI governance and fostering collaboration among stakeholders will be critical to harnessing AI's full potential while mitigating associated risks. Future research should focus on developing adaptive and ethical AI systems to ensure sustainable innovation in the financial industry.

Central Bank Digital Currencies (CBDCs) have emerged in response to accelerated digitalization, declining cash usage, the rise of cryptocurrencies, and the need to modernize payment systems. This paper examines the development and implementation of CBDCs, highlighting initiatives such as the Sand Dollar in the Bahamas, eNaira in Nigeria, and JAM-DEX in Jamaica, as well as prominent ongoing projects like the Digital Yuan, Digital Yen, Digital Dollar, Digital Euro, and Digital Pound. The research was made during October-december 2024 and employs methods such as literature review, bibliographic synthesis, and comparative analysis to assess the benefits, challenges, and current state of CBDC implementation. Key indicators analyzed include financial inclusion, transaction efficiency, cybersecurity, and the impact on monetary policies.The paper emphasizes the central role of central banks in fostering user trust and managing risks associated with financial innovation. It also underscores the importance of close collaboration between public and private sector actors to develop solutions tailored to the specific needs of individual economies. CBDCs are seen as a significant opportunity to reduce reliance on cash, combat money laundering and terrorism financing, and enhance transaction transparency.Findings highlight CBDCs’ contributions to financial inclusion and payment system modernization, as well as challenges related to privacy, security, and user acceptance. The paper concludes that CBDCs hold transformative potential for global financial systems but require innovative designs, international cooperation, and robust strategies to mitigate risks and maximize benefits. Future recommendations focus on global standardization, policy adaptation for effective implementation, and the exploration of emerging technologies to support widespread adoption.

Purpose: The objective of this paper is to evaluate the effectiveness of the M-score and F-score in detecting financial statement fraud in the Chinese market and to develop machine learning models tailored for detecting such fraudulent activities.

Design/Methodology/Approach: We utilize the data of fraudulent cases from the CSMAR database for the period 2010-2019 and implement a random sampling by industry to match between fraudulent enterprises and non-fraudulent enterprises. Based on this sample, we first test the effectiveness of M-score and F-score in detecting financial frauds among Chinese listed companies. Next, we construct the machine learning models—Random Forest, Gradient Boosting Decision Tree (GBDT), K-Nearest Neighbor (KNN) and Support Vector Machine (SVM)—using the constituent variables of F-score and M-score, along with an additional loss indicator. The performance of these models in detecting financial frauds is then comparatively assessed.

Findings: The results reveal varying degrees of ineffectiveness of the M-score and F-score in accurately identifying financially fraudulent companies in the Chinese market. In contrast, the machine learning models show satisfactory performance, each exhibiting distinct advantages in reducing false negative and false positive rates.

Practical Implications: This research presents effective machine learning models for detecting and predicting financial statement fraud in the Chinese context, helping investors mitigate risks associated with stock investments in the Chinese stock market.

We present a novel framework that integrates graph analytics with machine learning to assess the factors influencing the short-selling of publicly traded company shares. This approach centers on constructing a knowledge graph representing selected companies in the banking sector, along with their corporate and individual owners as nodes connected by weighted ownership relations. By extracting network-based features such as the PageRank centrality alongside traditional financial indicators like firm size, ownership concentration, and insider trading activities, this framework identifies and ranks the factors that drive short-selling behavior.

This study employs a regression analysis using models such as random forests, support vectors, and neural networks to quantify the relationship between these features and the short-selling position averages and standard deviations. Features like the largest shareholder’s stake and the Herfindahl–Hirschman Index (HHI) capture the concentration of ownership, while normalized insider trading data provide insights into market sentiment and stock volatility. A comparative analysis using the Shapley Additive Explanation (SHAP) values reveals that although liquidity-related measures are key predictors of the average short-selling positions, the ownership concentration and insider trading are also influential, especially in explaining fluctuations in short-selling activity.

Overall, these results underscore the transformative potential of combining a graph-based network analysis with machine learning techniques to enhance financial risk modeling and governance transparency. This integrated framework not only improves the detection of governance vulnerabilities but also offers valuable insights for regulators and investors. Future research could extend this approach to other sectors with complex ownership networks, further refining the predictive accuracy by incorporating real-time data and additional alternative data sources.

Assessing credit risk has become a key activity in risk management and is particularly relevant for lenders, investors, and overbuilding markets. The purpose of this study is to determine the extent to which new deep learning methods can change credit risk modelling with large data and algorithms because they may facilitate the performance of predictions and risk mitigation techniques. Using advanced neural networks such as CNNs and RNNs, this study analyzes authentication adequacy and default prediction models based on key borrower characteristics, their financial history, and relevant macroeconomic conditions. Deep learning models overcome the limitations of classical statistical methods and improve performance for much more complex tasks, such as classification and regression, in assessing credit risk. Furthermore, solutions to deep learning explanatory difficulties can be developed through the use of XAI methods. Such approaches make it possible for all stakeholders to utilize the results of the model, which, in turn, makes systems more transparent and trusted rather than using incomprehensible artificial intelligence. This study demonstrates how to allocate credit to optimize the default rate; in other words, it demonstrates how to build stronger financial systems. It expresses the significance of AI regarding the use of information within a changing business environment. This study facilitates emerging AI-driven finances by developing the underlying framework of credit risk analysis and its impact on the business world. In so doing, the paradigm of assessing credit risk is altered.

The adoption of generative AI technologies in the financial sector is transforming operational processes, decision-making, and customer interactions. While these innovations enhance efficiency, they also raise a critical question: how can financial institutions balance automation with the value of human expertise? This study proposes a novel framework categorizing applications of generative AI in finance along two dimensions: the degree of automation and the value added by human intervention. The framework, developed through a comprehensive literature review, is validated with case studies in areas such as portfolio management, compliance, and risk assessment. It categorizes applications into four quadrants, balancing low and high levels of automation and human expertise. The findings highlight the potential of hybrid models (Quadrant 4), where advanced automation is combined with human oversight, offering the greatest efficiency and accuracy. For instance, a fintech company implementing AI-driven compliance tools with human supervision enhanced error detection in regulatory filings while maintaining compliance standards. This research provides a structured framework for integrating generative AI into financial workflows, helping institutions optimize the balance between automation and human expertise. It offers practical insights for decision-makers and serves as a foundation for responsible AI adoption, ensuring operational efficiency and strategic soundness in an era of digital transformation.