Artificial Intelligence as a Service: The Evolving Conversation in State Taxation

By Geoff L. Gunnerson, Tony Caldwell, and Mark Hannah¹

Artificial Intelligence as a Service (AIaaS) represents the newest stage in cloud-based technology, offering businesses access to powerful machine-learning models and generative AI tools through subscription platforms. Similar to Software as a Service (SaaS), AIaaS allows users to access functionality remotely without taking ownership or control of the underlying software. Yet, the question of how states should tax AIaaS offerings remains unsettled. State departments of revenue are now encountering AIaaS offerings that do not fit comfortably within established tax frameworks, signaling the start of a broader rethinking of how AIaaS offerings should be categorized for tax purposes. For example, will state departments of revenue look at the “true object” or primary purpose of an AIaaS offering to determine its taxability? Further, will state departments of revenue consider whether the AIaaS offering is bundled with or separate from other taxable elements?

Background

Although AIaaS and SaaS share a delivery model, they differ in substance. SaaS delivers static functionality, as users access the same program each time they log in. AIaaS, by contrast, provides operational intelligence through systems that learn, adapt, and generate new outputs in response to user inputs and data. This capacity for continuous adaptation moves AIaaS beyond mere software access toward an evolving, service-based interaction between provider and customer. That difference, while technical in nature, may prove decisive for tax purposes because it alters the character of the transaction from the lease or use of software to the provision of an intelligent service.

State Level Developments

In 2025, both Illinois and Indiana issued rulings addressing the taxability of AI-based chatbot services, illustrating how states are beginning to confront these questions directly. In Illinois, the Illinois Department of Revenue² examined a company offering access to an AI-powered chatbot through its website and mobile app but analyzed the transaction through familiar SaaS principles. Because no software was transferred or owned, Illinois (at the state level) treated the service as nontaxable, a conclusion consistent with prior guidance on cloud-hosted software. Importantly, the result differs for AIaaS offerings provided to customers within the boundaries of Chicago, Illinois, which are subject to a different taxing regime.³

Indiana’s ruling on the subject struck a more explicit tone. The Indiana Department of Revenue⁴ confronted similar facts but spoke the language of AI directly, e.g., generative AI, training, deployment. It recognized that a company offering access to a chatbot involved machine-learning models built and refined on cloud infrastructure, an acknowledgment of AI as a distinct commercial service. Yet, when it came to classification, Indiana relied on SaaS-era definitions, finding the service nontaxable because customers lacked ownership or control of the underlying code. That tension, recognizing the novelty of AI while reasoning through inherited taxing frameworks, captures the moment we are in, namely, states are beginning to articulate what AI is, even as they define it by what it is not.

AIaaS has emerged within an uncertain tax landscape, where statutes designed for prewritten software struggle to capture the dynamics of learning-based technology. Even as this landscape takes shape, some states, such as Texas⁵ and Washington,⁶ already treat data-processing or digital-information services as taxable, suggesting that similar reasoning could apply to certain AI offerings. Other states like New York have proposed laws that would impose tax-like surcharges on corporations using AI to displace workers or for data mining.⁷ The result is a widening divide in how similar AI transactions may be characterized from one jurisdiction to the next. For taxpayers and practitioners, that divergence signals not only administrative inconsistency but also a deeper question of AI’s legal taxonomy and whether learning-based systems should be understood as property, service, or something entirely new.

Conclusions

State and local tax authorities will continue to revisit similar issues as AI becomes more embedded in business operations. Continued administrative and judicial developments will shape how AI is classified and taxed across jurisdictions. We will continue to monitor these developments and provide updates as new guidance emerges.