Skidmore Institutional Guidelines and Policy on the Usage of AI

I. Institutional Guidelines for the Use of AI

Purpose

AI technologies are rapidly shaping our world, and Skidmore has been proactive in understanding how these tools are used while weighing their potential impacts and ethical considerations. This document provides practical guidance and guardrails for the use of AI and AI-related tools at Skidmore College.

Our intent is to foster innovation while protecting institutional values, supporting creative and thoughtful work, and maintaining the human element in how we use these technologies.

Skidmore recognizes that experimentation with emerging technologies can support creativity, research, and effectiveness. Within these guardrails, members of the Skidmore community are encouraged to explore AI tools thoughtfully and responsibly, balancing innovation with attention to privacy, security, academic integrity, resource constraints, and institutional values.

Guiding Principles

• Users should approach AI as they would any other tool, exercising judgment, care, and accountability.
• AI use should be aligned with community expectations and institutional values.
• AI should be used in ways that support equity, accessibility, and inclusion.
• AI usage should support and enhance, not replace, human judgment, creativity, and accountability.

Responsible Use Expectations

• AI outputs should be reviewed and, where appropriate, validated for accuracy, appropriateness, and fairness before use.
• Users are responsible for verifying the validity of AI-assisted outputs and addressing potential bias.
• The use of AI should consider environmental and resource impacts, including computational intensity, energy consumption, and associated infrastructure demands.
• Where feasible, users should exercise judgment in selecting appropriately scaled tools for their task and avoid unnecessary or excessive use of resource-intensive systems.
• Skidmore recognizes that AI technologies have broader environmental and resource implications and will continue to evaluate these impacts as part of institutional decision-making around technology adoption and usage.
• Users should be mindful of how these tools may affect their well-being, focus, workload, boundaries, and interpersonal or community connection, and are encouraged to make use of Skidmore’s wellness resources as needed.

Institutional Focus Areas

This guidance and policy are centered on four key areas:

• Data privacy and protection
• Ethical use and impact
• Security and access control
• Training and awareness

Institutional Clarifications

• This document provides institutional guardrails while relying on existing governance structures.
• This document does not create new enforcement mechanisms.
• It does not define instructional or faculty policy. (See the CEPP statement on academic freedom and AI.) However, all uses of AI involving institutional systems, networks, platforms, or data remain subject to applicable institutional technology, security, and data governance requirements.
• It does not prescribe or require the use of AI tools.
• It does not replace broader data governance, academic integrity, or institutional policy frameworks.
• It aims to ensure the safe usage of tools and offer institutional guidance around their usage.

II. Institutional Policy for the Use of AI

Data Privacy and Protection

Data must be handled in accordance with Skidmore’s institutional data classification framework as defined by institutional data governance policies.

Public Data

• Data approved for public release and open distribution.
• May be used with approved or non-approved AI tools.

Internal Data – General (Level 1)

• Data broadly accessible for routine institutional operations.
• May be used within Approved/Sanctioned AI tools for institutional purposes.
• Users should exercise care to avoid unnecessary inclusion of personally identifiable or sensitive information.

Internal Data – Restricted (Level 2)

• Data requiring additional authorization due to legal, regulatory, or institutional requirements.
• May only be used within Approved/Sanctioned AI tools for approved institutional purposes.
• Requires appropriate authorization.

Internal Data – Confidential (Level 3)

• Highly sensitive data with significant risk if disclosed.
• Requires explicit approval from the CIO or designated authority prior to use in any AI tool.
• Must meet heightened security and access control requirements.
• Includes data that enables direct connection to institutional systems, platforms, or records through API, integration, or automated processes.

Security and Access Control

• AI tools must meet Skidmore’s security standards and comply with applicable laws and institutional policies.
• Skidmore IT will maintain an approved AI vendor registry.
• Non-approved tools may not be used with internal institutional data.
• AI tools and vendors must go through institutional procurement and security review processes prior to institutional use.
• Independent procurement of AI tools for institutional work should not be undertaken without consultation with IT.
• Storage or transmission of internal data through tools that do not meet institutional security requirements is prohibited.
• Users should remain vigilant to cybersecurity risks, including AI-enabled phishing, social engineering, and other forms of system compromise.
• Any data exposure (intentional or unintentional) must be reported to IT for assessment and response.

Ethical Use and Accountability

• Use of AI should be consistent with existing intellectual property, authorship, and academic integrity standards.
• Misuse of AI tools that results in harmful, biased, misleading, or inappropriate outcomes may be addressed through existing institutional policies and processes.
• Responsibility for AI-assisted work rests with the user.
• Consideration of environmental and resource impacts is part of responsible AI use and aligns with Skidmore’s broader institutional commitments.

Training and Awareness

• Individuals using AI tools are expected to complete approved training on responsible use, privacy, and bias mitigation.
• Ongoing training will be required as tools and policies evolve.
• This policy will be reviewed annually.

III. Applicability

This policy applies to all faculty, staff, and students who use AI tools on any college-owned devices or networks.

While the use of AI with public data carries minimal institutional risk, this policy is particularly relevant to the use of internal institutional data as defined by institutional data governance policies.

This policy supplements existing privacy laws and college policies (e.g., FERPA, HIPAA, GDPR), policies on academic integrity, and the Faculty Handbook.

IV. Responsible Office

Skidmore’s Information Technology office, in coordination with IPPC’s Subcommittee on Institutional Effectiveness and other relevant offices, is responsible for the administration, maintenance, and periodic review of this policy. This policy operates in alignment with existing institutional policies, including information technology and information security policies, which take precedence where applicable.

Appendix A: Definitions

Artificial Intelligence (AI)
Systems or software that perform tasks associated with human intelligence, including prediction, classification, recommendation, and content generation.

Generative AI (GenAI)
AI systems that generate new content (text, images, audio, video, or code) based on patterns learned from training data.

Large Language Model (LLM)
A type of generative AI model trained on large-scale text data to understand and produce human-like language.

AI Tool / AI Service
Any application, feature, or platform that uses AI or generative AI capabilities, including embedded AI features within existing software systems.

Approved/Sanctioned AI Tool
An AI tool that has been reviewed and approved by Skidmore IT and is included in an institutional registry of approved technologies.

Non-Approved AI Tool
Any AI tool that has not been reviewed or approved by Skidmore IT.

Data Classification

Data classifications referenced in this document align with institutional data governance policies and include:

• Public Data
• Internal Data – General (Level 1)
• Internal Data – Restricted (Level 2)
• Internal Data – Confidential (Level 3)

Approved by the Institutional Policy and Planning Committee May 8, 2026.