CFA Society Of Toronto - Home Page
In this article, we provide a summary of the 2025 CFA Institute report AI Washing: Signs, Symptoms, and Suggested Solutions for Investment Stakeholders by Joseph Simonian, with the aid of supplementary papers, including Order from Chaos: How Data Science Is Revolutionizing Investment Practice by Joseph Simonian, Marcos Lopez de Prado and Frank J. Fabozzi.
An increasing number of global financial institutions are using or considering using artificial intelligence (AI) for data analytics, trading, portfolio optimization, portfolio management, pricing, risk management and underwriting activities. Many have reported using generative AI to replicate data (synthetic financial time series, natural language processing, text and scenario generation) for practical applications. Enthusiasm and perceived benefits of AI applications, however, have led to a growing risk of “AI washing.” This is a risk to asset owners conducting manager search and due diligence. Therefore, it is essential that clients adopt the appropriate methodology to determine whether a firm’s claim of AI-driven
investment processes and products is genuine and value-added or merely commercially motivated and exaggerated. In the following sections, we define AI washing and its driving forces and develop a set of questionnaires to raise awareness about AI washing and provide guidance to institutions engaged in manager selection and evaluation.
What is AI and AI washing?
Simonian defines AI as a computational and/or statistical tool or system that exhibits an enhanced, extended or more effective replication of some well-defined aspects of human cognition. This is referred to as “weak AI.” This definition includes supervised and unsupervised machine learning, reinforcement learning, natural language processing and the various forms of so-called generative AI. “Strong AI,” in contrast, refers to a replication of generalized human intelligence, including emotions, common sense reasoning and a strong ability to contextualize.
AI washing refers to the use of buzzwords, marketing strategies and exaggerations of the true presence of AI in companies’ business activities, leading to client, regulator and investment community confusion, skepticism, distrust and potential ethical concerns.
“Investment practitioners are increasingly turning to data science and machine learning (ML) to enhance analytics, support decision-making and improve predictive modelling. Unlike traditional methods such as ordinary least squares (OLS), ML can process both structured and unstructured data and capture non-linear relationships in markets. Given the complexity of financial markets — shaped by human intentionality — ML and data science can address the idiosyncrasies in financial data, adding significant depth to investment research.”
The Role of Data Science in Finance and Investment Practice by Joseph Simonian, Marcos López de Prado and Frank J. Fabozzi
Why do companies engage in AI washing?
The risk of competitor firms moving faster to implement genuine machine learning and AI capabilities, as well as fear of missing out, can lead firms to engage in AI washing. The adverse effect of AI washing is compounded when investment firms are also reluctant to reveal proprietary processes and use the “secret sauce” defence when defining their methodologies. AI and machine learning algorithms can enhance risk management, trading and portfolio management. However, the practical implementation of AI in asset management presents additional challenges, as relevant data in investment management is typically more limited, more volatile and less uniform than in other areas of finance. An example of this challenge is predicting asset prices with far fewer observations, with drivers of asset behaviour often more complex than the drivers of other types of financially relevant behaviour. The proper application of AI in portfolio construction, risk management and trading requires focus, expertise and dedicated resources. Genuinely adopting new technologies and methodologies requires specialized talent and resources, which many firms struggle to procure due to costly and time-consuming hiring processes.
From a commercial perspective, adopting advanced technologies and methodologies can increase investment firms’ chances of attracting new business. At the same time, most professional investors already have well-developed investment processes that typically preclude them from making significant modifications to their existing procedures. For most mature quantitative firms, modifying their processes could risk potentially leading to unfavorable investment outcomes. To avoid paying a price for increased complexity, some engage in AI washing.
How can asset owners differentiate between legitimate AI technologies and inflated claims in the market?
A practical approach to evaluating the truthfulness of a firm’s AI claims is to inquire if the right personnel are working on AI projects. In the tech industry, leaders likely possess a great deal of expertise in AI, as the firm’s product is technology; otherwise, tech firms could negatively impact profitability. In contrast, the primary products of investment management firms are investment advice and asset management services, where the connection between technological innovation and product performance is less direct.
Asset managers must be able to provide sufficient details about why and how they implemented a piece of AI technology in their process, what specific frameworks they used and what results or improvements they observed. This aligns with the ethical principles of transparency and the duty to clients, as outlined in the CFA Institute Code of Ethics and Standards of Professional Conduct.
To guide due diligence, Simonian proposes a set of questions designed to assess the credibility of AI claims, presented below from AI Washing: Signs, Symptoms, and Suggested Solutions for Investment Stakeholders by Joseph Simonian:
• Can you specify what type of algorithm or combination of algorithms you are using and how it enhances the forecasting of asset returns?
• How does your AI-driven model outperform simpler models? Can you provide a quantitative comparison of relevant performance metrics?
Enthusiasm and perceived benefits of AI applications, however, have led to a growing risk of “AI washing.”
• What preprocessing and feature selection techniques are used to prepare the raw data for input into your model(s)? Do you use fundamental features, such as earnings surprise, price momentum or other signals and indicators? How do you preprocess the data before feeding the data into your model(s)? Do you standardize or normalize the input features, and what techniques do you use to handle missing data, outliers and limited datasets?
• How do you maximize model interpretability? Is it through model choice or post-implementation communication s? If the latter, can you give some concrete examples?
• Can you provide an example of a recent investment decision that was influenced by the model’s output? How was the rationale for that decision explained to the investment team?
• Can you provide out-of-sample backtest results or cross-validation results using your model? Have you tested the model’s efficacy on bootstrapped or otherwise synthetic data? How does the model perform relative to simpler models, traditional benchmarks and equal-weighted portfolios?
• How do you validate the robustness of the models you develop? What precautions do you take to guard against overfitting? For example, how do you tune hyperparameters into your models? How do you monitor “model drift,” and what mechanisms are in place to retrain the models and/or adapt to shifts in the market landscape?
• What governance structures are in place to ensure the responsible use of AI firmwide? Do you have an internal AI audit process, and how often are the models reviewed for compliance with generally accepted standards and protocols?
and robustness of the services and products used in your investment process?
Protecting investors and safeguarding trust
Ultimately, mitigating AI washing is about protecting investors and safeguarding trust in the financial industry’s adoption of emerging technologies. Financial firms selling financial products should adhere to the same standards of transparency that stakeholders, regulators, customers and the investment community demand from other investment products, and must refrain from exaggerating the use of AI technology and its success in the investment process. By asking relevant questions and embedding such scrutiny into the manager selection process, stakeholders, particularly customers and regulators, can not only minimize the risk of AI washing but also foster genuine innovation and trust in financial markets, ultimately achieving better investment outcomes.
The price of persistence 
Assessing management quality in public markets: What to look for in corporate leadership 
Private markets: The recent landscape and thoughts on what comes next