AI rebuilding a shattered financial building

Can AI Predict Bankruptcy? How New Models Are Changing Finance

Mira Elwood in Business & Economy December 2025 • 4 min read.

"Explore how multimodal AI models are revolutionizing bankruptcy prediction by analyzing financial texts, market data, and accounting information, offering a more accurate and comprehensive risk assessment."

Predicting which companies will face bankruptcy has always been a critical task for investors, regulators, and the financial industry. Traditional methods often rely on analyzing financial statements and market trends, but these approaches can be limited and sometimes fail to capture the full picture of a company's financial health.

Now, artificial intelligence (AI) is stepping in to revolutionize bankruptcy prediction. New AI models are capable of processing vast amounts of data, including financial filings, market data, and even textual information, to provide more accurate and timely risk assessments. This article explores how these AI models work and what impact they could have on the future of finance.

One of the most promising advancements is the use of multimodal AI models, which combine different types of data to gain a more comprehensive understanding of a company's financial situation. By integrating textual data from financial filings with traditional numerical data, these models can offer a more nuanced and forward-looking view of bankruptcy risk.

The Power of Multimodal AI in Predicting Bankruptcy

AI rebuilding a shattered financial building

Traditional bankruptcy prediction models primarily use accounting and market data. While these are valuable indicators, they often lack the forward-looking insights that can be gleaned from textual data. Financial filings, such as the Management's Discussion & Analysis (MDA) section in Form 10-K, contain critical information about a company's strategies, challenges, and future prospects.

However, accessing and processing this textual data can be challenging. Not all companies are required to submit an MDA, and technical difficulties can arise when trying to extract the information. This is where multimodal AI models come in. These models can learn from accounting, market, and textual data, creating a more holistic representation of a company's financial health.

Comprehensive Data Analysis: Multimodal AI models analyze accounting data, market trends, and textual information from financial filings.
Improved Accuracy: By combining diverse data sources, these models provide more accurate bankruptcy predictions than traditional methods.
Forward-Looking Insights: Textual data offers a forward-looking view of a company's prospects, helping to identify risks early on.
Overcoming Data Limitations: AI can fill in data gaps and overcome technical challenges in extracting textual information.

One example of such a model is the Conditional Multimodal Discriminative (CMMD) model. The CMMD model learns multimodal representations that embed information from accounting, market, and textual data. During training, the model needs a sample with all data modalities. At test time, however, it only needs access to accounting and market data to generate multimodal representations, which are then used to make bankruptcy predictions and even generate words from the missing MDA modality.

The Future of Bankruptcy Prediction

Multimodal AI models represent a significant step forward in bankruptcy prediction. By combining diverse data sources and leveraging advanced machine learning techniques, these models offer a more accurate, timely, and comprehensive assessment of financial risk. As AI continues to evolve, it is likely to play an increasingly important role in helping investors, regulators, and companies navigate the complex world of finance.

About this Article -

This article was crafted using a human-AI hybrid and collaborative approach. AI assisted our team with initial drafting, research insights, identifying key questions, and image generation. Our human editors guided topic selection, defined the angle, structured the content, ensured factual accuracy and relevance, refined the tone, and conducted thorough editing to deliver helpful, high-quality information.See our About page for more information.

This article is based on research published under:

DOI-LINK: https://doi.org/10.48550/arXiv.2211.08405,

Title: Multimodal Generative Models For Bankruptcy Prediction Using Textual Data

Subject: q-fin.rm cs.lg stat.ml

Authors: Rogelio A. Mancisidor, Kjersti Aas

Published: 26-10-2022

Everything You Need To Know

What are the limitations of traditional methods for predicting bankruptcy?

Traditional bankruptcy prediction methods typically rely on analyzing financial statements and market trends. While valuable, these approaches often lack the forward-looking insights needed to fully assess a company's financial health. They may not capture the nuances present in textual data, such as discussions of strategies, challenges, and future prospects found in financial filings like the Management's Discussion & Analysis (MDA) section of Form 10-K.

How do multimodal AI models improve bankruptcy prediction?

Multimodal AI models enhance bankruptcy prediction by integrating diverse data sources, including accounting data, market trends, and textual information from financial filings. This comprehensive approach allows the models to provide more accurate and timely risk assessments compared to traditional methods. By combining these data types, multimodal AI models gain a more nuanced and forward-looking view of a company's financial situation.

What is the role of textual data in multimodal AI models for bankruptcy prediction, and what challenges exist in using it?

Textual data, particularly from financial filings like the Management's Discussion & Analysis (MDA) section in Form 10-K, offers forward-looking insights into a company's strategies, challenges, and future prospects. However, accessing and processing this data can be challenging. Not all companies are required to submit an MDA, and technical difficulties may arise when extracting the information. Multimodal AI models address these challenges by learning from accounting, market, and textual data to create a more holistic representation of a company's financial health.

Can you explain how the Conditional Multimodal Discriminative (CMMD) model works in predicting bankruptcy?

The Conditional Multimodal Discriminative (CMMD) model learns multimodal representations by embedding information from accounting, market, and textual data. During training, the model requires a sample with all data modalities. However, at test time, it only needs access to accounting and market data to generate multimodal representations. These representations are then used to make bankruptcy predictions and even generate words from the missing MDA modality, showcasing its ability to leverage available data effectively.

What are the broader implications of using AI, particularly multimodal AI models, in the financial industry for predicting bankruptcy?

The use of AI, specifically multimodal AI models, in bankruptcy prediction represents a significant advancement in the financial industry. These models offer a more accurate, timely, and comprehensive assessment of financial risk by combining diverse data sources and leveraging advanced machine-learning techniques. As AI continues to evolve, it is likely to play an increasingly important role in helping investors, regulators, and companies navigate the complex world of finance, enabling earlier and more reliable detection of financial distress. The insights gained can improve decision-making and risk management across various sectors.