uConnect's Responsible AI Approach

 What Responsible AI Means at uConnect

uConnect's approach to AI is grounded in four principles: accuracy, privacy, transparency, and human oversight. These aren't aspirational values — they're design constraints that shape how AI Search is built and maintained.

Accuracy: AI Search Only Says What It Knows

AI Search is designed to prioritize accuracy over comprehensiveness. When it can't find a relevant answer in your VCC content, it says so — clearly and without fabrication. The system never generates a confident-sounding response for a question it cannot actually answer from retrieved content.

This design choice — declining to answer rather than making something up — is a deliberate responsible AI decision. A system that admits its limitations is more trustworthy than one that always produces a response.

Privacy: Your Data Stays Yours

• Tenant isolation — Your institution's content is strictly separated from all other institutions on the platform. Students at one school will never see content from another, and your institutional data is never accessible to other uConnect clients.
• No open internet — AI Search operates exclusively within your VCC content. Students receive only institution-vetted, approved information — eliminating the risk of unverified, harmful, or biased content from the web.
• No cross-institutional data sharing — Student activity data is not shared across institutions.

Transparency: Guardrails as a Feature

AI Search is built with explicit scope boundaries — known as guardrails — that define what questions it will and won't attempt to answer. These guardrails cover:

• Out-of-scope topics — questions about housing, admissions, health services, or other non-career domains are gracefully declined, with the student redirected to the appropriate resource.
• Out-of-scope coaching — AI Search will not provide personalized career advice, resume rewrites, or emotional support. It surfaces resources; it does not replace human advisors.
• Prohibited or sensitive queries — the system is built to recognize and respond appropriately to queries that fall outside appropriate boundaries.

uConnect treats guardrails as a product feature, not just a safety fallback. Well-designed boundaries reinforce what the tool is best at and help students find their footing quickly when they've asked something outside its scope.

Human Oversight: The Feedback Loop

AI Search is not a fully autonomous system. It is actively monitored and improved by uConnect's product and engineering teams. The thumbs up / thumbs down feedback that students provide is reviewed by real people and used to tune the system's performance over time.

This human-in-the-loop philosophy means that as AI Search gets more usage, it gets better — and that improvement is guided by actual student feedback, not just automated metrics.

Security Compliance

• SOC 2 compliance — uConnect maintains SOC 2 compliance with rigorous governance frameworks and continuous monitoring across all platform infrastructure, including AI Search.
• HECVAT participation — uConnect participates in HECVAT (Higher Education Community Vendor Assessment Toolkit) security assessments. If your institution's IT or procurement team requires a HECVAT response, contact your CSM.