{"id":33481,"date":"2026-08-21T11:12:48","date_gmt":"2026-08-21T15:12:48","guid":{"rendered":"https:\/\/www.uvm.edu\/it\/kb\/?post_type=ht_kb&p=33481"},"modified":"2026-09-25T11:47:21","modified_gmt":"2026-09-25T15:47:21","slug":"agentic-ai-dos-and-do-nots","status":"publish","type":"ht_kb","link":"https:\/\/www.uvm.edu\/it\/kb\/article\/agentic-ai-dos-and-do-nots\/","title":{"rendered":"Agentic AI Dos and Do Nots"},"content":{"rendered":" \t\t
\r\n \t\t\tEarly Release Content<\/span> \t\t\t \t\t\t\t

\r\n \t\t\t\t\tThis is an early release of this KB article and content is subject to change. Please refer back to this article occasionally for updated guidance. \t\t\t\t<\/p>\r\n \t\t\t \t\t\t\r\n \t\t<\/div>\r\n \t\t\n \t\t

\r\n \t\t\tData Classifications in AI<\/span> \t\t\t \t\t\t\t

\r\n \t\t\t\t\tSelect AI tools are UVM are only to be used with Low, Moderate, and High data types. Restricted data types are prohibited from being entered into any AI tool. A list of approved tools and associated data types can be found here on our Supported AI Tools page<\/a>. \t\t\t\t<\/p>\r\n \t\t\t \t\t\t\r\n \t\t<\/div>\r\n \t\t\n

Agentic AI tools can help you complete complex tasks, interact with applications, and take actions on your behalf. These capabilities can support innovation and productivity, but they also require careful oversight. The following dos and don\u2019ts provide practical guidance for using Agentic AI responsibly while protecting University data, systems, and people.<\/p>\n

Dos<\/span>\u00a0<\/span><\/h2>\n

Do treat AI agents like\u00a0they have your system access<\/strong><\/p>\n

By\u00a0default,\u00a0AI\u00a0agents use your\u00a0permissions and\u00a0user account. Make sure\u00a0all actions taken\u00a0by\u00a0the\u00a0agent\u00a0are\u00a0intentional and\u00a0monitored.<\/span><\/p>\n

Do enforce least privilege access<\/strong><\/p>\n

Restrict the access of the agent so it can only use the data, applications, and systems required. This is often referred to as least privilege access and is important for protecting you and UVM.<\/span><\/p>\n

Do protect sensitive data<\/strong><\/p>\n

Use anonymized\u00a0or non-sensitive data unless explicitly approved\u00a0by the Information Security Office and\/or Contract Review\u00a0for protected\u00a0data\u00a0(e.g., NPPD, research data).\u00a0Please refer to\u00a0UVMs\u00a0article on\u00a0<\/span>AI privacy<\/span><\/a>\u00a0for more information.<\/span><\/p>\n

Do implement logging and monitoring<\/strong><\/p>\n

Capture\u00a0and\u00a0monitor\u00a0agent\u00a0actions to other\u00a0applications,\u00a0data\u00a0sources,\u00a0and\u00a0decisions. Actions\u00a0should be\u00a0monitored\u00a0for unexpected behavior such as\u00a0access to data you\u00a0didn\u2019t\u00a0specify\u00a0or\u00a0to\u00a0the\u00a0agent\u00a0sending the data\u00a0elsewhere.<\/span><\/p>\n

Do require human oversight for high-impact actions<\/strong><\/p>\n

Keep humans in the loop for\u00a0actions involving\u00a0data\u00a0(especially modifications or deletions),\u00a0other applications, or external communications.<\/span><\/p>\n

Do consult UVMs approved list of\u00a0AI\u00a0tools<\/strong><\/p>\n

The approved list of tools can be found\u00a0<\/span>here<\/span><\/a>.\u00a0Do make sure you understand the\u00a0<\/span>UVM Data Classification Matrix<\/span><\/a>\u00a0before putting any work or research data into an AI\u00a0tool.<\/span><\/p>\n

Do Nots<\/span><\/h2>\n

Do not expose confidential or regulated data to unapproved AI tools<\/strong><\/p>\n

Only university approved tools can provide you with robust data protection agreements. Using unapproved tools, or tools managed with a personal account, do not provide you with robust data protection agreements.<\/span><\/p>\n

Do not grant broad or persistent access<\/strong><\/p>\n

Agent permissions and access should be granted on an as-needed basis for only as long as required. Giving an agent persistent and robust permissions expose you to data related risks.<\/span><\/p>\n

Do not bypass security controls<\/strong><\/p>\n

Neither you or your agent may modify or bypass authentication, authorization, or monitoring mechanisms.<\/span><\/p>\n

Do not assume AI interactions are private<\/strong><\/p>\n

While UVM offers robust data protection for AI tools including agents, all vendors maintain the right to retain and review prompts and outputs for security reasons. Do not assume anything provided to a UVM approved AI tool is anonymous.<\/span><\/p>\n

Do not allow fully autonomous operation for high-risk tasks<\/strong><\/p>\n

Do not allow your agent to perform any unsupervised actions involving sensitive data, financial actions, or system changes. All agent actions taken on your behalf are your responsibility. For that reason, you need always be aware of what your agent is doing.<\/span><\/p>\n

Do not allow AI systems to make autonomous decisions about humans<\/strong><\/p>\n

If an AI\u00a0agent\u00a0is\u00a0producing outputs that could affect, or are related\u00a0to,\u00a0a\u00a0human,\u00a0you must\u00a0supervise the\u00a0actions taken\u00a0and\u00a0review\u00a0output. AI\u00a0agents and systems\u00a0cannot be held accountable for actions\u00a0taken,\u00a0so you must review anything produced by them.<\/span><\/p>\n

Do not assume everything your agent reads is safe<\/strong><\/p>\n

Content from web pages, documents, or remote services can\u00a0contain\u00a0hidden instructions that your agent will act on. Be especially careful when working with outside data sources.<\/span><\/p>\n","protected":false},"excerpt":{"rendered":"

Agentic AI tools can help you complete complex tasks, interact with applications, and take actions on your behalf. These capabilities can support innovation and productivity, but they also require careful oversight. The following dos and don\u2019ts provide practical guidance for using Agentic AI responsibly while protecting University data, systems, and…<\/p>\n","protected":false},"author":117,"comment_status":"closed","ping_status":"closed","template":"","format":"standard","meta":{"_relevanssi_hide_post":"","_relevanssi_hide_content":"","_relevanssi_pin_for_all":"","_relevanssi_pin_keywords":"","_relevanssi_unpin_keywords":"","_relevanssi_related_keywords":"","_relevanssi_related_include_ids":"","_relevanssi_related_exclude_ids":"","_relevanssi_related_no_append":"","_relevanssi_related_not_related":"","_relevanssi_related_posts":"","_relevanssi_noindex_reason":"","_monsterinsights_skip_tracking":false,"footnotes":""},"ht-kb-category":[662],"ht-kb-tag":[],"class_list":["post-33481","ht_kb","type-ht_kb","status-publish","format-standard","hentry","ht_kb_category-ai"],"_links":{"self":[{"href":"https:\/\/www.uvm.edu\/it\/kb\/wp-json\/wp\/v2\/ht-kb\/33481","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.uvm.edu\/it\/kb\/wp-json\/wp\/v2\/ht-kb"}],"about":[{"href":"https:\/\/www.uvm.edu\/it\/kb\/wp-json\/wp\/v2\/types\/ht_kb"}],"author":[{"embeddable":true,"href":"https:\/\/www.uvm.edu\/it\/kb\/wp-json\/wp\/v2\/users\/117"}],"replies":[{"embeddable":true,"href":"https:\/\/www.uvm.edu\/it\/kb\/wp-json\/wp\/v2\/comments?post=33481"}],"version-history":[{"count":7,"href":"https:\/\/www.uvm.edu\/it\/kb\/wp-json\/wp\/v2\/ht-kb\/33481\/revisions"}],"predecessor-version":[{"id":33748,"href":"https:\/\/www.uvm.edu\/it\/kb\/wp-json\/wp\/v2\/ht-kb\/33481\/revisions\/33748"}],"wp:attachment":[{"href":"https:\/\/www.uvm.edu\/it\/kb\/wp-json\/wp\/v2\/media?parent=33481"}],"wp:term":[{"taxonomy":"ht_kb_category","embeddable":true,"href":"https:\/\/www.uvm.edu\/it\/kb\/wp-json\/wp\/v2\/ht-kb-category?post=33481"},{"taxonomy":"ht_kb_tag","embeddable":true,"href":"https:\/\/www.uvm.edu\/it\/kb\/wp-json\/wp\/v2\/ht-kb-tag?post=33481"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}