Executive summary. As artificial intelligence (AI) spreads into commercial, academic, and business applications, problems will inevitably arise. What are these problems, and which stakeholders are affected? Moreover, how do corporate managers and educators prepare employees and students for these issues? These incidents are common and meaningful. For example, Rite Aid’s facial recognition technology allegedly classified a disproportionate number of minorities and women as shoplifters. Yet, there’s been little to examine these problems systematically. We study 225 harmful AI incidents involving public companies from 2013 to 2023. These incidents reveal significant issues not just with technology but also with social, ethical, and safety implications.
Key takeaways
- AI incidents harm shareholder value, in addition to creating social and ethical problems.
- Better corporate governance mitigates the losses from AI problems.
- Companies respond with increased risk management.
- Business leaders, policy makers, and academic institutions must educate their students and workforce not only for AI integration but also for AI crisis management.
Introduction
Concern over AI adoption is ubiquitous. In industry, output from the ‘black box’ of AI processing must be checked for errors or unintended outcomes. Even Meta’s security researcher, Summer Yue, lost a chunk of her email inbox when her AI agent decided to follow its own rules. She had specifically instructed it not to delete anything without her permission (Martindale, 2026). Despite the negative consequences of incidents, AI has the potential to improve product innovation, corporate productivity, and labour output (Babina et al., 2024; Brynjolfsson et al., 2025). So, it is crucial to overcome problems and concerns with AI adoption.
These concerns about AI in the corporate setting spill over, directly and indirectly, into academia, and there is increasing focus on both the output and use of AI in higher education, as well as a recognition that students must learn about AI to integrate into the future workforce. While education often focuses on AI integration, an ostensibly understudied aspect of AI adoption is crisis and risk management, since AI will undoubtedly malfunction some of the time.
Within higher education, AI problems affect both researchers and students. AI hallucinations lead to citations of non-existent works, false facts, and ideas lacking internal consistency. Yet academics and students increasingly utilise AI tools to support their research. Pedagogically, business students use AI to look up concepts, complete assignments, write reports, and (sometimes unscrupulously) answer exam questions.
As in the corporate setting, most people in academia recognise the importance of understanding and developing AI-related skills for use in universities and in the corporate world after graduation. AI is likely to dominate in future business activities. These tasks include analysing corporate decisions, devising portfolio investment strategies, communicating with customers, and developing skills for future careers as managers or investors.
It is not only important to help educate students to develop an understanding of how to use AI effectively to enhance productivity, but it will also be necessary to help students recognise AI’s limitations and how to overcome problems when it is wrong, inconsistent, or flat-out misbehaving. Hence, the future of education must not only include edification about the use and benefits of AI, but also examples and understanding about AI’s issues and how to minimise their negative consequences.
When AI is Wrong, What Are the Consequences?
A casual inspection of recent AI problems reveals that they extend beyond simple technological issues, such as unauthorised email deletion. The potential for problems goes much deeper. A couple of incidents at Amazon highlight how AI issues are spreading into the social and ethical realms. In one case, Amazon’s technology had a “cataloguing error” that led to LGBT-themed books losing their rankings and visibility on its sales platform. In a separate Amazon-related incident, their AI recruiting tool appeared to downrank female applicants. In short, AI can introduce issues related to gender, safety, and equity in the corporate environment. Focusing on these potential downsides could help ensure thoughtful adoption of AI technologies and minimise frictions.
AI Incidents Research
In our recent working paper (Jiang, Neyland, Wang, and Zhang, 2026), we examine AI incidents between 2013 and 2023 from incidentdatabase.ai, which aggregates complaints and details about large language models, artificial intelligence, and related technologies (‘AI incidents’). We uncovered a dramatic increase in AI incidents, especially since 2020. While AI has been improving in quality and accuracy, its more widespread adoption has led to greater problems overall.
To better understand the determinants and impact of AI incidents in the business and finance industry, we focus on corporate incidents and classify them into five categories based on their characteristics: automation safety, AI-generated information quality, governance of online media, privacy, and employment fairness. While incidents related to information quality have continued to increase, incidents involving automation safety and online media governance have decreased significantly (see Figure 1.)
These incidents, as expected, lead to losses in company value, as seen in stock price declines. Stock price drops (‘abnormal returns’) accumulate to 8.67% over 60 days around the incidents. This effect is not uniform. While most incidents are associated with losses to shareholders, those involving employment fairness are surprisingly associated with price increases averaging 8% to 14%. That is, shareholders may positively value AI’s bias in employment decisions!
Management Communication with Shareholders
We next examine how shareholders respond to firms’ disclosures of potential AI incidents and risks. In financial disclosures (10-K’s), firms are required to disclose significant risks as part of Item 1A. We find that share prices do not react significantly to incidents when managers have previously disclosed potential risks related to artificial intelligence and machine learning. This lack of reaction suggests candid management communication can help prime investors for these incidents, reassure shareholders of management’s consideration of the problem, and potentially, limit the losses that stakeholders ultimately face.
How Do Firms React to AI Problems?
Managers must handle the backlash to problems associated with AI incidents. Whether it’s a driverless car crash, the removal of legitimate social media posts, or the generation of fictitious, if not comical, hallucinations about real events, companies must work to prevent the recurrence of these problems. Externally, there is a fledgling industry that provides assurance services for AI and related technologies. Internally, one would expect managers to change some operations.
We have collected firms’ employment data for risk management-related positions on LinkedIn from 2014 to 2017, and patent data to assess how firms react to incidents. Companies expand their risk management and product quality positions following incidents, consistent with efforts to mitigate their negative effects and prevent future incidents. They also slow down their innovation activities, as evidenced by a reduction in patenting in the years following the incidents.
The Societal Reaction
Given the benefits to society and the need to reduce incidents, government action is imminent. Product failures or output biases have already drawn attention from enforcement agencies and plaintiffs (D’Acunto et al., 2019; Florackis et al., 2023). At the legislative level, concerns have led to laws requiring the publication of safety frameworks and the disclosure of serious safety incidents, such as California law SB 53.
What Does the Future Hold for AI Problems?
Further AI incidents are certain. The primary questions revolve around how many problems to risk in the name of technological progress. While AI-related incidents and problems are likely to decrease (per use) as technology progresses, current trends point to increased technological and social problems as AI is increasingly incorporated into our businesses, universities, and lives. Looking to the near future, there are likely to be more problems and opportunities stemming from AI incidents and the efforts required to prevent and address them.
It is incumbent on educators to prepare students for this future and both the opportunities and challenges it poses. This will require discussion, collaboration, and implementation of AI-related topics and skill-building in higher education. Given the uncertain direction of new technologies, specific mechanisms to guide the curriculum’s future remain unclear. However, necessary elements will likely include connections between academics and industry. This will ensure that pedagogy focuses on the latest industry trends and developments and that students are better prepared for the business environment and the needs of future employers.
References
Babina, T., Fedyk, A., He, A., Hodson, J. (2024). Artificial intelligence, firm growth, and product innovation. Journal of Financial Economics (151), 103745.
Brynjolfsson, E., Li, D., Raymond, L. (2025). Generative AI at work. The Quarterly Journal of Economics (140), pp. 889-942.
D’Acunto, F., Prabhala, N., Rossi, A. G. (2019). The promises and pitfalls of robo-advising. The Review of Financial Studies (32), pp. 1983-2020.
Florackis, C., Louca, C., Michaely, R., Weber, M. (2023). Cybersecurity risk. The Review of Financial Studies (36) pp. 351-407.
Jiang, T., Neyland, J., Wang, Z., and Zhang, F. (2026). AI Incidents. Working paper.
Martindale, Jon. Meta Security Researcher’s AI Agent Accidentally Deleted Her Emails. PC MAG (February 24, 2026). Retrieved at
https://www.pcmag.com/news/meta-security-researchers-openclaw-ai-agent-accidentally-deleted-her-emails
Tianqi Jiang, PhD, is an Assistant Professor of Finance at Beijing Jiaotong University and serves as the Associate Director of the Career Guidance Centre. Professor Jiang’s research focuses on corporate finance, securities analysts, and fintech, and she received a PhD in Finance from the University of Rhode Island.
Jordan Neyland, JD, PhD, is an Associate Professor of Finance at Bentley and formerly an Assistant Professor of Law at the Scalia Law School. His work has been published in the Journal of Accounting & Economics and the Journal of Corporation Law, among other journals in law, finance, and accounting.
Zhao Wang, PhD, is an Associate Professor of Finance at the Capital University of Economics and Business. Professor Wang’s research spans corporate governance, disclosure, and fintech and has been cited in Bloomberg. He received a PhD in Finance from the University of Rhode Island.
Fan Zhang, PhD, is an Assistant Professor of Finance at Bentley University and has research interests focusing on corporate finance, political economy, and institutional investors. She has a PhD in Mathematics from the University of Miami and in Finance from Arizona State University.
