The study examines how academic libraries are approaching the licensing of AI applications. The findings show a market still very much in transition. Only 38.46% of surveyed institutions report that they currently lease or license generative AI tools for their academic communities, while an identical 38.46% say they do not, and 23.08% report that they are planning to do so. Adoption varies sharply by institution type: just 20.00% of CC/BA-level colleges currently license AI tools, compared with 50.00% of doctoral universities and 44.44% of research universities. MA-level colleges stand out for hesitation, with 66.67% reporting no current licensing and 0.00% planning to do so.
Vendor-specific leasing remains limited. Only 23.08% of respondents reported licensing ChatGPT Edu or Enterprise, while 76.92% said they did not. Among institutions above 8,500 FTE, however, adoption rose to 50.00%, highlighting the role of scale and resources. By contrast, none of the smallest enrollment bands reported licensing enterprise AI platforms.
Governance and risk issues loom large. When asked about xAI Grok for academic use, 61.54% of respondents said the product had not been evaluated at all, and another 30.77% judged it not suitable or not very suitable—meaning nearly 89% expressed no positive assessment. Safety concerns, reputational risk, and data-governance issues were cited most often. More generally, 76.92% of respondents said vendors had not agreed to adjust safety filters for legitimate research use.
Across questions, libraries emphasized that cost, FERPA compliance, privacy protections, and whether institutional data feeds model training matter more than raw model performance. Many respondents reported enterprise quotes “in the six figures,” limited pricing transparency, and uncertainty created by AI features being bundled into existing platforms.