diff --git a/content/_policies/artificial-intelligence.md b/content/_policies/artificial-intelligence.md
index cd4e4c6..565a696 100644
--- a/content/_policies/artificial-intelligence.md
+++ b/content/_policies/artificial-intelligence.md
@@ -24,4 +24,7 @@ active use and must release and maintain that code as open source software on a
 OMB memorandum [M-24-18](https://www.whitehouse.gov/wp-content/uploads/2024/10/M-24-18-AI-Acquisition-Memorandum.pdf) further extends the provisions of M-24-10 but focuses on the acquisition of AI technology. It reiterates the requirement from the [Federal Source Code Policy](/info/software-development/#open-source) for agencies to include language in software development contracts which states that any vendor-produced code must be open source. It also introduces new requirements for contracts with AI vendors requiring data management practices, and disclosures as to the use of copyrighted materials in training data, as this could put the government at significant legal risk.
 {:#M-24-18}
 
-Notably, this memo *does not* cover the use of open source models which have not been purchased.
\ No newline at end of file
+Notably, this memo *does not* cover the use of open source models which have not been purchased.
+
+@WARNING It is important to note that almost all of the most popular AI models have End-User License Agreements (EULAs) and/or Terms of Service (TOS) that the federal government cannot legally agree to, as they contain provisions requiring indemnification (a violation of the Anti-Deficiency Act ([31 USC 1341](https://uscode.house.gov/view.xhtml?path=/prelim@title31/subtitle2/chapter13/subchapter3&edition=prelim)), agreeing to arbitration, and other restrictions. Even for open source models, the EULAs and TOS are agreed to simply by being used and would be illegal. It is strongly suggested that agencies consult their general counsels, and negotiate individual terms with their vendors before using any AI models.
+{:.notes}
\ No newline at end of file