<div dir="ltr"><p style="margin-right:0cm;margin-left:0cm;font-size:12pt;font-family:"Times New Roman",serif"><span style="font-family:Cambria,serif;color:black">The findings show a
market still very much in transition. Only <strong>38.46%</strong> of surveyed institutions report that
they currently lease or license generative AI tools for their academic
communities, while an identical <strong>38.46%</strong>
say they do not, and <strong>23.08%</strong>
report that they are planning to do so. Adoption varies sharply by institution
type: just <strong>20.00%</strong>
of CC/BA‑level colleges currently license AI tools, compared with <strong>50.00%</strong> of doctoral
universities and <strong>44.44%</strong>
of research universities. MA‑level colleges stand out for hesitation, with <strong>66.67%</strong> reporting no current
licensing and <strong>0.00%</strong>
planning to do so.</span></p>

<p style="margin-right:0cm;margin-left:0cm;font-size:12pt;font-family:"Times New Roman",serif"><span style="font-family:Cambria,serif;color:black">Vendor‑specific
leasing remains limited. Only <strong>23.08%</strong>
of respondents reported licensing ChatGPT Edu or Enterprise, while <strong>76.92%</strong> said they did not.
Among institutions above<b> <strong>8,500
FTE</strong></b>, however, adoption rose to <strong>50.00%</strong><b>,</b> highlighting
the role of scale and resources. By contrast, none of the smallest enrollment
bands reported licensing enterprise AI platforms.</span></p>

<p style="margin-right:0cm;margin-left:0cm;font-size:12pt;font-family:"Times New Roman",serif"><span style="font-family:Cambria,serif;color:black">Governance and risk
issues loom large. When asked about xAI Grok for academic use, <strong>61.54%</strong> of respondents said
the product had not been evaluated at all, and another <strong>30.77%</strong> judged it not
suitable or not very suitable—meaning nearly <strong>89%</strong> expressed no positive assessment. Safety
concerns, reputational risk, and data‑governance issues were cited most often.
More generally, <strong>76.92%</strong>
of respondents said vendors had not agreed to adjust safety filters for
legitimate research use.</span></p>

<p class="MsoNormal" style="margin:0cm;font-size:12pt;font-family:Cambria,serif"><span style="color:black">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.</span></p>

<p class="MsoNormal" style="margin:0cm;font-size:12pt;font-family:Cambria,serif"> </p>

<p class="MsoNormal" style="margin:0cm;font-size:12pt;font-family:Cambria,serif">For further information view our website at <a href="http://www.PrimaryResearch.com">www.PrimaryResearch.com</a></p>

<p class="MsoNormal" style="margin:0cm;font-size:12pt;font-family:Cambria,serif"> </p>

<p class="MsoNormal" style="margin:0cm;font-size:12pt;font-family:Cambria,serif"><b>James Moses, Research Director </b></p>

<p class="MsoNormal" style="margin:0cm;font-size:12pt;font-family:Cambria,serif">Primary Research Group Inc. </p>

<p class="MsoNormal" style="margin:0cm;font-size:12pt;font-family:Cambria,serif"><a href="mailto:jmoses@primaryresearch.com" style="color:rgb(5,99,193)">jmoses@primaryresearch.com</a>
</p>

<p class="MsoNormal" style="margin:0cm;font-size:12pt;font-family:Cambria,serif"><a href="http://www.PrimaryResearch.com">www.PrimaryResearch.com</a></p></div>