Building responsible universities in the age of AI

Key risks of AI use in higher education

The growing presence of AI tools is changing how knowledge is produced and evaluated. Findings from the Digital Education Council’s Global AI Faculty Survey 2025 and the FICCI-EY-Parthenon’s AI Adoption in Higher Education Survey 2025 reflect concerns about misuse and about students losing essential critical thinking skills in AI-mediated learning environments. These learning risks highlight the need for clearer AI policies for universities and stronger governance across higher education institutions. 

A significant area of concern is the use of AI classroom tools for grading and feedback. Automated scoring systems, especially those trained on standardized writing patterns, can unintentionally disadvantage students whose linguistic or cultural backgrounds differ from dominant norms. This makes ensuring equity in AI-based grading and evaluation a critical institutional responsibility. Meanwhile, uneven faculty familiarity with AI tools leads to unclear expectations, inconsistent classroom usage, and confusion among students about what constitutes ethical AI use in academia.

Blanket restrictions on AI are neither realistic nor aligned with preparing learners for AI‑enabled professional environments. As a result, universities are rethinking assessment design. Many are adopting strategies that address academic challenges created by AI‑led learning, using academic assessment redesign approaches that emphasize real-time performance rather than polished outputs easily generated by AI systems.

 Academic integrity and assessment redesign

Institutions are increasingly adopting three broad assessment categories, including:

• AI‑free assessments rely on formats like supervised exams, in‑class writing tasks, and oral examinations to test unaided student thinking.

• AI‑assisted assessments allow controlled use of tools, requiring students to document how AI contributed to their work.

• AI‑integrated assessments embed AI tools directly into learning activities—treating them as partners for ideation and feedback—while evaluating students’ reasoning, judgment and conceptual mastery.

A notable example comes from the NYU Stern School of Business, where an AI-enabled oral exam was piloted in a Product Management course. An AI agent posed questions, analyzed responses, and adapted follow‑ups to probe deeper understanding. The recorded interactions were later evaluated using AI‑supported grading to enhance consistency. This case demonstrates how AI assessment frameworks can evolve responsibly without compromising academic standards.

Rather than restricting AI use in written work, faculty are increasingly redesigning assessments to function within an AI-rich environment. Formats such as oral examinations and in-class problem solving test real-time understanding and reduce over-reliance on automated tools. This transition reflects a broader institutional shift toward responsible AI framework design.

Multi‑stakeholder governance frameworks

Effective AI governance in universities extends beyond classrooms. Because AI influences teaching, research, administration, admissions and student‑facing services, universities must adopt governance models that cut across functions.

• Faculty uphold academic standards and develop discipline‑specific guidelines.
• Students, as primary AI users, need consistency, transparency, and protection from opaque algorithmic decisions.
• IT and compliance teams ensure cybersecurity, system integrity, and strong student data privacy protocols.
• Institutional leadership oversees strategy, accountability, and regulatory compliance.

These responsibilities underscore why responsible AI governance frameworks for universities must be collaborative and multi‑layered. Many institutions are forming central AI councils to develop principles and review emerging risks. Others are adopting federated models, setting shared guardrails while allowing flexibility for different disciplines. Regardless of the structure, governance must include mechanisms for monitoring, revision, and adaptation as technology evolves.

Regulatory considerations and data protection

In India, responsible AI adoption must comply with the Digital Personal Data Protection Act (DPDP Act), notified in November 2025. Educational institutions are classified as Data Fiduciaries, which impose specific obligations around lawful data collection, processing, retention, and sharing. All data use must be purpose‑specific and supported by informed consent.

These requirements significantly shape how universities deploy AI systems. Legacy datasets created for administrative or instructional reasons cannot automatically be repurposed for analytics or AI model training. If such secondary uses were not previously disclosed, institutions may need new consent from data principals. Cross-border data transfers remain possible but may face restrictions or localisation requirements. With compliance monitored by the Data Protection Board, universities must treat data governance as a central pillar of AI strategy.

Because AI systems rely heavily on data flows, alignment with the DPDP Act could shape both the pace and scope of adoption. Robust data protection in educational practices safeguards against innovation compromising privacy, trust, or institutional credibility.