USA Trends and Developments Contributed by: Jeffrey Harvey, Randall Parks, Andrew Geyer and Cecilia Oh, Hunton Andrews Kurth LLP
providers are even willing to guarantee some level of savings based on relatively opaque application of these “back-office” efficiencies gained through AI. As a result, buyers of technology are more likely to pro - cure generative AI and agentic AI solutions on a one- off basis, often to address a small handful of internal use cases, rather than as part of a larger outsourcing transaction. The legal issues raised as a result of the provision and use of these new technologies are not entirely new and usually revolve around the following: • ownership of intellectual property in the bots (or, in the case of generative and agentic AI, the “learn - ings” and the outputs); • pricing of additional bots (both new development and cloning); • avoiding proprietary automation platform lock-in; • privacy and infringement concerns over AI tools “scraping” the internet; • biased data (or biased human intervention in the
have already introduced or passed legislation aimed at improving transparency and establishing account - ability standards to curb such misuse. For example, in 2025, Nebraska enacted the Ensuring Transparency in Prior Authorization Act, pursuant to which utilisation review agents are prohibited from relying solely on AI- based algorithms to deny, delay or modify healthcare services based on medical necessity. In 2025, Maine adopted An Act to Ensure Transparency in Consumer Transactions Involving Artificial Intelligence, which prohibits the use of AI chatbots or other computer technology to engage in commercial transactions with consumers that may mislead or deceive a consumer into believing that they are engaging with a human being, unless the consumer is notified in a clear and conspicuous manner that they are not engaging with a human being. In 2024, Colorado passed the Colorado Artificial Intelligence Act, which will allow employees to challenge a private company’s decision not to hire them if AI was used as part of the decision-making process. In 2023, New York adopted a state law man - dating bias audits for AI tools used in employment decision-making, covering tools used for hiring and promotion decisions. In 2019, Illinois adopted the Arti - ficial Intelligence Video Interview Act, which prohibits an Illinois employer from using AI to evaluate job inter - view videos in certain circumstances and, in particular, places an emphasis on the potential for racial biases resulting from the use of AI. Similar bills have been introduced or enacted in Colorado, California, Massa - chusetts, Maryland, New Jersey, Washington and New York City, some of which would impose bias audit - ing and other compliance requirements on AI users, enforced through civil penalties. Additionally, multiple states have enacted AI-targeted amendments to their respective privacy laws. Colorado, Connecticut, Utah and Virginia, for example, have enacted laws that (i) give consumers the right to opt out of automated pro - filing and (ii) require a data protection assessment for activities that pose a “heightened risk of harm”. In the 2023 legislative session, Indiana, Montana, Oregon, Tennessee and Texas also passed consumer privacy laws which include provisions governing AI, including some that mirror those passed by Colorado, Con - necticut and Virginia. As of July 2023, the National Conference of State Leg - islature was tracking legislation addressing AI in all 50
data) used to develop AI models; • data protection and ownership; • sharing of savings; and • displacement of workers.
As the proliferation of agentic AI models increases, concerns over the limits and autonomy of these mod - els will also increase. Machine learning and AI Machine learning and AI solutions are capable of sorting through massive amounts of data in order to, in many cases, reach their own conclusions. Absent human intervention, there is no room for context or consideration of “soft” factors, and the solutions reach conclusions based solely on the data they were trained on and subsequently collect. This one-track mindedness of the solutions poses problems when the output is integrated into decision-making process - es that carry the potential for legal liability. Legislators and regulators have taken notice of the potential for misuse of AI with encoded bias – such as discriminatory outcomes in hiring, healthcare and law enforcement – and the growing concern that AI tools can pose as real human beings, and states
106 CHAMBERS.COM
Powered by FlippingBook