Since its launch in November 2022, numerous authorities have raised concerns about the risks posed by OpenAI’s ChatGPT technology. In March, Italy’s Data Protection Agency took the extraordinary step of banning ChatGPT within the country over concerns about consent and personal data privacy.
Ironically, this one-month ban may have provided the strongest evidence to date of the technology’s transformative impact on business and the economy.
Capitalizing on this rare natural experiment opportunity, researchers at Olin Business School at Washington University in St. Louis and the National University of Singapore examined changes in the relative stock market valuation of Italian firms across multiple industries during the one-month ban to assess the ban’s consequences and generative AI’s overall value.
Their findings — published in a working paper, “Capital Market Consequences of Generative AI: Early Evidence from the Ban of ChatGPT in Italy,” — show the ban’s impact was twofold: Not only did the ban negatively impact firm productivity, it also impacted investor behavior.
“Not since the introduction of the internet has a technology so quickly transformed how businesses operate and compete. What’s remarkable is that it’s easy and cheap to adopt, helping small, underdog businesses compete with much larger firms without making heavy investments in infrastructure or human capital,” said Jeremy Bertomeu, an associate professor of accounting at Olin Business School.
Measuring ban’s impact on businesses
Across the globe, businesses have leveraged AI generative technology to optimize processes that previously required significant labor and infrastructure support. Not surprisingly, the researchers found early adopters — including firms that provide professional, scientific and technological services — were hardest hit by Italy’s temporary ban.
According to the data, Italian firms in these high-exposure industries experienced an average negative return of 8.7% compared with similar European stocks during the monthlong ban.
They also found evidence that the ban had a greater impact on newer and small Italian businesses. Compared with larger, more established Italian businesses, these firms underperformed by 6.8-7.1%, respectively.
“The negative market reaction indicates that new, small and tech-savvy businesses benefit the most from generative AI because it allows them to reduce the information advantages held by larger incumbent firms and narrow the competitive gap.”
How the ban impacted investors, businesses’ bottom line
Because the goal of the research was to understand the capital market consequences of the ban, it was also important to assess how the ban influenced investor behavior.
“ChatGPT transforms the investment landscape by providing investors, especially small investors, with a conversational and interactive platform to gather insights and navigate the complexities of the financial market,” Bertomeu said.
The research showed that the loss of this investment tool led to an increase in information asymmetry during the ban. The effect was greatest for firms with fewer institutional investors, limited analyst coverage and a lower presence of foreign investors — which would still have access to AI technology during Italy’s ban.
As a result of this information asymmetry, bid-ask spreads — the difference between the highest price a buyer will offer and the lowest price a seller will accept — widened during the ban and firms suffered decreased liquidity.
“In the EU, it is unlikely that other countries will follow suit with a ban, but regulators in France, Germany, Spain, among others, met to discuss whether AI complies with EU privacy laws, and this could lead to future restrictions on the technology,” Bertomeu said.
A cautionary tale
In the short time since its launch, generative AI technology like ChatGPT has revolutionized businesses worldwide, providing a powerful tool for innovation and creation. Its influence on businesses worldwide will only continue to grow.
Yet, simultaneously, governments worldwide are grappling with potential security threats and ethical concerns related to technology. Bertomeu hopes the early evidence presented in this case study will offer a cautionary example of the potential consequences of regulating AI.
“No one asked the Italian regulatory agency to ban ChatGPT, but they did it anyway without any consultation of affected parties or elected officials,” Bertomeu said. “Regulatory policy should always be based on careful cost-benefit analyses and public input. Our data demonstrate what can go wrong when regulators skip these fundamental steps.”
Bertomeu’s co-authors include Yibin Liu, Yupeng Lin and Zhenghui Ni, all from the National University of Singapore.