Emma Bailey
James Parsons

As this blog’s readers know, AI is not only an emerging technological phenomenon it is also a potentially disruptive source of D&O risk and liability. In the following guest post, Emma Bailey and James Parsons, take a look at the contours of the developing AI-related D&O risk and discuss the implications. Emma is Senior Underwriter, Commercial Management Liability, Berkshire Hathaway Specialty Insurance, London, and James is Senior Claims Examiner, Executive & Professional Liability, Berkshire Hathaway Specialty Insurance, London. I would like to thank James and Emma for allowing me to publish their article as a guest post on this site. I welcome guest post submissions from responsible authors on topics of interest to this site’s readers. Please contact me directly if you would like to submit a guest post. Here is the authors’ article.

***********************

As underwriters, we’re constantly evaluating emerging risks—and few topics are as dynamic and widely debated as Artificial Intelligence (AI). It’s impossible to reduce a projected $1 trillion+ in AI-related capital expenditure to a simple yes or no answer, however, in this article we will explore some AI terminology, how we’re approaching AI as a risk factor, and what impact it is—and isn’t—having on Directors & Officers (D&O) litigation, particularly for US-listed insureds.

What is AI?

What appears to be newer to the risk register is a concept that dates back several decades. However, more recent and accessible use has propelled the broad term ‘Artificial Intelligence’ into the public eye, even in the most traditional of industries. AI can be described as an area of computer science and engineering that focuses on developing software capable of solving problems which have historically required human ability to think and reason, with speed and effectiveness, to reach a conclusion or a decision. Artificial Intelligence is a multidisciplinary field which is often used interchangeably with other terms such as Machine Learning, Deep Learning, and Neural Networks.

MACHINE LEARNING – An older form of AI (existing since the 1950’s) which consists of algorithms that allow computers to learn without human involvement.

DEEP LEARNING – An advancement of Machine Learning that uses the neural networks to perform human-like tasks involving judgement and reason.

NEURAL NETWORKS – Inspired by the human brain, this describes different collections of algorithms focused on different areas of data, which is processed in order to learn but all the areas are interconnected.

PREDICTIVE AI – Refers to the way you can use machine learning to study data to find patterns or trends within the data. The computer can then use what it has found to make predictions on future patterns without a human having to program it.

GENERATIVE AI – Using deep learning and neural networks to respond to language prompts with newly generated data which could be in the form of text, images or other.

Is AI a disruptive technology?

The term is being used in public fillings, regulations, news and related media – and in my opinion, it’s rather good at designing a two-week holiday itinerary. Globally around 78% of organisations are said to have used Gen AI in at least one business function, with 42% using it for marketing and sales purposes, and 28% for product and/or service development.

Despite this, adoption still looks to be in early stages, with most products that use Gen AI remaining in pilot phase. As a result, there are a number of, and sometimes conflicting schools of thought, supported by corresponding literature surrounding the future of the technology. This raises important questions for businesses, investors, regulators and other stakeholders:

  • Projected capex: Will the cost of adoption (be it investment, recruitment, servicing, integration or employee acceptance of change) translate into a directly better long-term business valuation?
  • Impact: How can you accurately measure efficiency gains that came directly from Gen AI usage – and will this be done consistently across public company disclosures?
  • Competition: What is the unique selling point if any potential margin impact or revenue increase can also be replicated by competitors?
  • Regulation: What and how will the industry be regulated?
  • Environmental Impact: Is the global grid capable of keeping up with the electricity power that is needed and are there political and ethical changes required to get there?
  • Strategy: What is the balance between being a first mover and learning from the mistakes of others?

Do we see AI as a risk factor for every customer?

It is likely there are some industries that will be impacted or disrupted to a greater or lesser extent than others. As we are still early in the cycle, it is hard to pinpoint who the winners or losers will be. Previous experience with disruptive technologies suggests that being first past the post doesn’t necessarily guarantee success.

However, what is clear is the increasing mention of AI related disclosures on public earnings calls. According to FactSet, more than 210 or 40% of S&P 500 Companies have cited “AI” on earnings calls in Q1 2025, this being the fifth consecutive quarter. This significantly exceeds the 5-year average of 114 and the 10-year average of 72.

This trend is not limited to US companies. Similar analysis by Bloomberg found that mentions of “AI” on Q4 2024 earnings calls from companies in the Stoxx Europe 600 index reached a record 1,788, around 20% more than the prior year.

General concerns raised by AI exposure and the potential for related litigation include ‘AI Washing’ which refers to a deceptive marketing strategy in which a company overstates their product or service offerings around AI. Some companies have also been accused of trying to capitalise on this by exaggeration of capabilities. More recently, we have seen allegations of under disclosure regarding business model disruption due to broader adoption of AI.

Privacy and security also raise concerns. Some customers may not want their risk information or data used in the development or training of AI systems, and will the AI capabilities of one company comply with their customers by-laws, codes of conduct and business agreements, particularly in a global economy across jurisdictions, regulations and cultures. Other areas of interest include the assumptions used in providing quarterly and annual guidance along with Risk Factors and disclosures in public filings. Increasingly, attention is being paid to the tone of management discussions in fillings, particularly as some readers are using AI supported screening tools to quickly compare company results and review the strength of the language used6. Further concerns regarding the implementation of AI include the extent to which a company’s board delegate tasks to AI, and how will these decisions leave D&O’s liable for those when there may not be a statement made by an individual.

A VIEW FROM CLAIMS

What impact is AI having on D&O litigation?

The focus of this litigation analysis is on US listed companies, which to date is where we have seen the most D&O litigation. According to the chart ‘Federal and State Class Action Filings Summary’ as depicted in figure 1 from Cornerstone’s 2024 Year In Review, SCA’s (core filings, exc. M&A) filed in 2024 increased to 225, from 215 in 2023, and from 208 in 2022, and was 14% higher than the historical average. The majority of these filings (95%) are rule 10b-5 claims, which is the highest level in more than 5 years, and roughly similar to the average level from 1997-2023. The industries which are impacted the most are the services, technology, financial and pharmaceutical companies.

SCA’s are on the rise (increase of 10) and AI-related litigation doubled from 7 to 15 in 2024, so it appears this is a new focus of underlying allegations from the Plaintiff Bar, but this could simply be a shift of focus into the relative unknown, where all parties (companies, plaintiff’s, insurers) are learning what to disclose and how AI will impact their business models. What we must also learn in time as insurers, is how AI related SCA filings are comparable to the average SCA filings.

Cornerstone Research found that filing activity in the first half of the year has remained on par with the second half of 2024, with 114 filings compared to 115, respectively. 12 class actions filed in the AI category “When annualized, the number of AI-related filings is on pace to far surpass the 2024 total (15),” Whilst potentially unrelated to the increase in specific AI-related SCA filings, what is interesting is that the Maximum Dollar Loss (MDL) sharply increased to nearly USD 1.9 trillion in H1 2025 from USD 730 billion in H2 2024 which could be an indicator of an increased level of severity in settlements.

It is too early to determine at this stage, but these AI related filings could have a higher dismissal rate or severity rate which could cause the Plaintiff Bar to really focus on AI disclosures as a source of litigation. In the early days of AI litigation, the allegations related to overstating AI capabilities, failing to disclose AI risks, and misleading investors about AI business models. As generative AI use becomes more widely accepted, allegations will continue to evolve too as plaintiffs are continually tracking and challenging business behaviours. Trends in litigation are constantly changing; we have seen COVID, ESG and Crypto related SCA’s all be a focus with various degrees of success, and it appears AI is the current focus. These new trends of litigation, shift over time with a number of factors behind them, such as;

  • Macroeconomic factors such as inflation and interest rates, global supply chain issues, foreign exchange markets and stock markets.
  • Political regimes – In the US, federal AI regulation has not developed at the pace that many expected. The current administration has issued an Executive Order emphasizing the goal of maintaining the US at the forefront of AI innovation. The appointment of Paul Atkins as SEC Chair may influence how the SEC balances the removal of barriers to innovation with the need for regulatory oversight in a competitive global marketplace.
  • Industry specific factors – AI is more prevalent in certain industries.

Of all the AI related Securities Class Actions that have been filed (as at August 1, 2025), there have been 50 complaints according to our records which in some way relate to AI; this could be against an AI company, or AI is referenced in the complaint with varying degrees of materiality. Of the 50, only 15 directly relate to AI washing, i.e. AI specific disclosures made by the company, and that is the alleged reason for the stock drop and misrepresentations. Of those, one has been voluntarily dismissed, and the remainder are pending – it remains to be seen how successful these pending matters will develop. At BHSI, we will continue to closely monitor new and current AI-related SCA’s as they proceed through the litigation and settlement process, to compare against aggregate SCA’s. The success, or lack of, will be a huge driver in determining if the Plaintiff Bar continue to focus on AI related SCA’s. Until we start to see some decisions, particularly at the Motion to Dismiss stage, the plaintiffs’ bar is free to capitalize on the “unknown” and the overall concern companies have as to the impact of AI on business models and the level of disclosure needed.

Based on Cornerstone Research SCA Filings 2024 Year In Review, the graph ‘Summary of Trend Filings – Core Federal Filings’, this shows the various different branches of allegations in SCA’s in the years 2020 – 2024. Whilst the AI litigation has indeed increased, it seems to have essentially off set the decreases in other previous allegation options.

We have also seen a decline in Environmental related cases. According to Nera, “there were five environment- related securities class action suits filed in 2024, a 38% decline from the eight cases seen in 2023.” This further demonstrates that AI-related and AI-driven litigation is a shift from a previous hot topic to a new mode of litigation.

Conclusion

What we’ve observed over the past few years is growing awareness of AI and its capabilities from both business leaders and consumers. The discussion is broad and has many angles but with AI references in earnings transcripts consistently exceeding five-year averages and the industry mix of those mentions expanding beyond Technology, we believe the topic is here to stay. So, while the current impact on claims trends involves a shift in the theme behind allegations, what remains to be seen is what the long-term severity impact of AI related SCAs with 50 open cases. With respect to any settlement data, the lack of decisions (either dismissals or not) means that we cannot yet determine the impact these pending SCA’s will have on losses to the D&O market, and if any will follow the traditional patterns of settlement. As the AI space evolves, and the current and new cases mature, it could be a natural development to experience more litigation focused around this area, but only if the Plaintiff bar also experience a similar relative success in bringing these suits.