Considerations and Use of AI
Ethical considerations describe the broader risks that AI introduces for the College and the communities it serves. They are areas of ongoing attention that help faculty, staff, and students judge when and how to use AI, and to recognize consequences that may not be visible in everyday use.
Ethical Considerations
Data Privacy and Consent
Information entered into an AI tool, including prompts, uploaded documents, and usage patterns, is transmitted to the vendor and handled under its terms of service. Depending on the vendor, that information may be retained for extended periods, used to train future models, or shared with third parties, and the terms themselves can change over time. Standard consent mechanisms give users limited insight into these practices, and information incorporated into training data is difficult or impossible to remove afterward. Prompts can also contain personal information about other people, such as students, colleagues, or community members, whose consent was never sought.
Accuracy and Misinformation
AI models generate text by predicting plausible language from statistical patterns, which means outputs can be fluent, confident, and factually incorrect at the same time. Documented failure modes include fabricated citations, invented statistics, and misattributed quotations, and the quality of the writing provides no indication of the quality of the underlying information. Errors spread when unverified output is circulated or cited as though it had been sourced. Generated images, audio, and video raise related concerns, since synthetic media can convincingly depict people and events that do not exist.
Indigenous Knowledge, Data Sovereignty, and Cultural Harm
Training data describes Indigenous Peoples largely through material written about them rather than by them, and AI outputs reflect this. Documented problems include pan-Indigenous portrayals that blur distinct nations, cultural practices presented inaccurately or out of context, and teachings wrongly attributed to Elders and Knowledge Keepers. A separate consideration is data sovereignty: Indigenous communities hold governance over their own data, stories, languages, and knowledge, and that authority is not set aside by public availability. Entering Indigenous knowledge into an AI system, or circulating AI-generated content about Indigenous Peoples, can therefore breach community protocols and cause cultural harm.
Environmental Impact
AI systems run on data center infrastructure that consumes electricity and water, and demand is growing as adoption spreads. The effects are most direct in the regions that host the facilities, where impacts on electricity prices, water supplies, and noise levels have been documented. The footprint of any individual interaction is small, but institutional adoption aggregates many interactions, and public reporting from the industry on energy and water use remains limited.
Unequal Access to Tools
Access to AI tools is uneven: cost, connectivity, language, and disability affect who can use them effectively. This becomes a consideration wherever AI use is built into academic or professional expectations, since those who can't access or effectively use the tools may be disadvantaged relative to those who can.
Bias
AI systems learn from training data that contains existing societal biases, including patterns tied to race, gender, disability, and socioeconomic status, and can reproduce those patterns in their outputs. Because outputs are fluent and appear neutral, biased content can be difficult to detect, and the consequences are most significant where outputs inform assessments or decisions about people.
Intellectual Property and Copyright
Copyright considerations apply to both the inputs and the outputs of AI tools. Uploading protected material, such as textbooks, journal articles, or another person's slides or writing, involves reproducing that work, and the licenses covering institutional materials often do not extend to this use. On the output side, models are trained on large bodies of existing work and can produce content that closely resembles protected sources, while ownership of AI-generated content remains unsettled in Canadian law. Work that incorporates AI output can therefore carry intellectual property risk that is not apparent from the content itself.
Dependency and Erosion of Critical Skills
Cognitive offloading refers to delegating mental work to an external tool rather than performing it directly. AI extends the practice across a wide range of academic tasks, including reading, analysis, structuring an argument, and writing. Early research associates habitual offloading with weaker critical thinking, shallower engagement with material, and reduced retention, which matters in an educational setting because these tasks are how the underlying skills develop. Sustained reliance carries similar implications for professional judgment among faculty and staff.
Responsible Use
Responsible use sets the expectations for individual conduct: the working habits that answer the ethical considerations AI brings. These practices apply to everyone at the College and operate alongside existing policies on academic integrity, privacy, and acceptable use of technology.
Disclose and Attribute AI Use
When AI contributes to a piece of work, through drafting, research, summarization, or editing, say so. Follow the College's citation and attribution guidelines, keep a brief record of how AI was used for anything beyond routine drafting, and where expectations are unclear, ask before proceeding and lean toward disclosure.
For students, this means disclosing AI use in coursework under the rules of each course; submitting undisclosed AI work where it is not permitted is academic misconduct. For faculty, it means being open with students when AI helps shape course materials or feedback. For staff, it means acknowledging AI assistance in reports and institutional communications.
Verify All AI Outputs
You are accountable for the accuracy and quality of any AI-assisted work you produce or share. Fact-check claims, review for bias and gaps, and apply your own judgment before an output goes anywhere, because errors reflect on the author regardless of what produced them. AI must not be the basis for decisions that affect individuals, such as grading, admissions, hiring, or discipline, without meaningful human review.
For students, AI-assisted work is assessed to the same standard as any other work. For faculty, verification extends to AI-generated feedback and course materials before students see them, and to AI detection tools, whose results are unreliable and are not evidence of misconduct on their own; concerns belong in existing academic integrity processes. For staff, AI-assisted work passes through normal supervisory review, and decisions informed by it should be documented.
Protect Personal and Confidential Information
Treat every prompt as a disclosure. Before entering any information into an AI tool, apply the principle of least use: provide the minimum the task actually requires and leave out anything the tool does not need to produce a useful answer. Work only with data you have the authority to use, which means information about students, colleagues, or community members enters a tool only where a legitimate College purpose and the legal authority to handle that information are both in place. De-identify wherever possible by removing names, identification numbers, and details that could single a person out, since a de-identified prompt usually serves the task just as well. Where doubt remains about whether information belongs in an AI tool, keep it out and ask, and route personal, confidential, or sensitive institutional information only through tools the College has approved for that purpose.
Respect Intellectual Property
Intellectual property deserves attention on both sides of an AI exchange: what goes in and what comes out. On the input side, submit only material you own or are authorized to use. Textbooks, journal articles, slide decks, images, and the work of colleagues and classmates are protected by copyright(opens in new window) even when they are easy to copy, so confirm permission or license terms before uploading work created by someone else. On the output side, generated content may reproduce protected material and its ownership is unsettled, so review outputs before publishing and cite AI contributions in accordance with applicable guidelines.
Use AI Purposefully
AI use carries financial, environmental, and cognitive costs, so reach for it when it adds real value rather than out of habit. The environmental cost tracks your choices directly: output format drives footprint far more than frequency, so favour text over generated images and video where either would serve, and think through a prompt rather than iterating reflexively. Ask whether a task genuinely benefits from AI at all, or whether working without it would better serve your learning, professional development, or creative goals.
Respect Indigenous Knowledge and Protocols
Indigenous data, stories, languages, and knowledge enter an AI tool only with the express authorization of the community concerned, and that authorization is required even for material that circulates publicly, because communities hold governance over their knowledge wherever it travels. Before any material connected to Indigenous Peoples goes into an AI system, confirm the legal and contractual requirements that apply, the privacy and confidentiality obligations involved, and the position of the community that holds authority over the material. Take particular care when generating text or images relating to Indigenous Peoples, cultures, or histories, since AI output in this area misrepresents routinely, and consult the appropriate Indigenous voices, whether colleagues, advisors, or community contacts, before anything generated is posted or shared. When researching Indigenous topics, rely on Indigenous-led and community-authorized sources, and treat AI-generated content about Indigenous Peoples as a draft requiring verification.
Commit to Ongoing Learning
The AI landscape moves quickly, and staying informed is part of using it responsibly. Treat that learning as a shared effort: bring what you discover to colleagues and classmates, and draw on what they have found in turn. Learn from the full range of sources you trusted before AI arrived, including books, journals, courses, practitioners, and communities of practice, in whatever form they take. Institutional guidance will continue to be revised as the technology and its implications become better understood.