Inside the 2025 UVA Conference on Ethical AI in Business

As generative AI’s capabilities and pitfalls continue to accelerate, the University of Virginia convened faculty, technologists, economists and even fiction writers at The Forum Hotel on the Darden School of Business Grounds to explore “minding the gap” between AI’s promise and shortcomings.

Darden Dean Scott Beardsley opened the day by introducing the LaCross Institute for Ethical Artificial Intelligence in Business, which hosted the conference and aims to equip leaders with the critical insights and ethical grounding needed to thrive in the AI-driven world. Beardsley reminded attendees just how economically significant AI has already become.

“If you just take the top six or seven AI companies… their valuations have been roughly $16 trillion,” a number he estimates as equivalent to the GDP of the 100 lowest-income countries in the world.

But he cautioned that business value alone doesn’t answer the harder questions: What are sustainable business models? Who benefits? And what ethical dilemmas emerge when incentives collide with societal well-being? These are among the questions the LaCross Institute explores.

Minding the Gap

The conference theme, “Minding the Gap,” was introduced by LaCross Institute Director Marc Ruggiano and Darden professor Raj Venkatesan. “Whenever we talk about the opportunities [in AI], there are also the challenges” of bias, misinformation, access and inequality, said Venkatesan. “We built this conference around the question: how do we build the bridges between the possibilities and the challenges?”

AI in the Real World: Evidence from Anthropic

The first keynote featured Peter McCrory, Head of Economics at Anthropic, who researches what AI technology like Anthropic’s Claude might mean for the economy and tracks how these models are actually being used in the real world.

“Our big motivation here is that AI is poised to have large but uncertain effects on the economy,” he said. Adoption is fast, capabilities are improving at startling speed, yet economic predictions vary wildly—from negligible impact to exponential productivity.

Anthropic’s research shows that the tasks users give to large language models like Claude currently map mostly to software development, followed by writing, research and other knowledge-work activities. “Three or four in 10 conversations… are people using Claude to help them with software development,” McCrory reported, versus only 3–4% of U.S. workers in that sector.

In the research, McCrory studies not only what people use Claude for, but how they use it. He breaks it down into two categories:

• Augmentation, where people collaborate with the model
• Automation, where users delegate tasks entirely

For the first time, McCrory said, “this automation category exceeds augmentation” on Anthropic’s platform—a trend that may signal shifting labor dynamics ahead. He notes that although we are still at the beginning of businesses adopting these tools, as AI capabilities are used more in business operations, the change may be invisible to consumers.

“Much like when I go to a coffee shop and I buy a latte, I don’t think about the electricity nor the infrastructure required to supply the electricity to the espresso maker to make my drink,” he said. “That is where the productivity and labor market effects will begin to concentrate as businesses begin to deploy these tools.”

What happens next depends largely on societal choices, he said. “The economic effects of transformative AI will be shaped as much by technical capabilities as by the policy choices that societies make.”

Darden ethicist Bobby Parmar spoke with UVA English professor and novelist Bruce Holsinger over lunch.

Ethical Dilemmas

Most of the day featured breakout sessions with subject matter experts exploring variations on the “minding the gap” theme, such as issues surrounding AI and privacy, investing, operations, entrepreneurship and more.

In a session on digital twins of populations moderated by Darden professor Tim Laseter, UVA computer science professor Anil Vullikanti demonstrated how high-fidelity simulations of social systems can enable disease forecasting, disaster planning and other public-impact use cases—but also raise complex privacy issues.

While his study showed that it is possible to implement privacy techniques on the data collected to create the model and still come out with improved results, it doesn’t always work. Maintaining privacy while improving the model is still an unsolved challenge.

A standout lunchtime conversation matched Darden ethicist Bobby Parmar with UVA English professor and novelist Bruce Holsinger, whose new book Culpability explores questions of guilt and responsibility after a family in a self-driving car is involved in a fatal car crash. Holsinger says he chose the world “culpability” as the title to invoke “that kind of guilt that all of humanity shares…everyone is responsible, not only everyone, but in this case everything in that car that day, including the AI that’s ‘driving.’”

AGI or Bust

The final keynote came from economist and Darden professor Anton Korinek, who has become one of the leading voices on AI-driven macroeconomic disruption.

“From an economic perspective…the resulting shift is going to be fundamental. I believe it’s going to be as fundamental as the industrial revolution,” Korinek said.

He noted that the goal of frontier AI labs is to develop AGI – artificial general intelligence, which is defined by OpenAI as an AI system that can perform all economically valuable tasks. If that happens, it raises many challenges in the long term such as inequality, income distribution and challenges to our systems of education.

Attendees meet and greet in between sessions at The Forum Hotel in Charlottesville, Va.

“Disruption is guaranteed,” he says of the short term. “The tech sector is placing an increasingly leveraged bet on AGI or bust. If the bet on AGI succeeds, we will see huge disruption in the labor market, and it’s going to be disruptive to the very fabric of our society. And if the bet on AGI fails, we are going to be seeing the collapse of a massive financial bubble.”

Yael Grushka-Cockayne closed out the conference, thanking speakers and attendees and summing up the day.

“Existing AI capabilities are already transforming the world around us,” she said. “We know that…but early AI adoption is uneven. The world is fascinated by what the appropriate use of AI is, and what the long-term implications will be. It’s still an open question.”