Why Trustworthy AI Matters More Than Ever | Faculty Fridays with Professor James Pang

Why It Matters

Trustworthy AI safeguards data privacy and decision integrity, enabling businesses to harness powerful models without exposing sensitive information.

Key Takeaways

• Trustworthy AI must be safe, secure, transparent, and value‑aligned.
• Unreliable AI can cause bias, unfairness, and harmful outcomes.
• Companies fear leaking confidential data when using public models like ChatGPT.
• The “confusion prompt” framework splits queries and adds fake questions.
• Assembled answers protect privacy while preserving original query’s usefulness.

Summary

In a Faculty Fridays interview, Professor James Pang outlines the growing urgency of building trustworthy artificial‑intelligence systems that users can rely on. He defines trustworthy AI as safe, secure, transparent, and aligned with human values, emphasizing that everyday decisions increasingly depend on these technologies.

Pang warns that without trustworthiness, AI can produce biased, unfair, or even harmful outcomes, jeopardizing both individuals and broader society. He highlights a concrete business concern: firms using public models such as ChatGPT fear inadvertent exposure of proprietary or personal data.

To address this, Pang’s team developed a “confusion prompt” framework. The method decomposes a user’s original query into multiple sub‑questions, injects fabricated distractor questions, sends the batch to the AI, and then algorithmically reconstructs the answer to the real query—preserving utility while shielding sensitive information.

The approach signals a shift toward privacy‑preserving AI deployments, offering enterprises a practical tool to mitigate data‑leakage risk and comply with emerging regulations, thereby accelerating responsible AI adoption across industries.

Original Description

As AI becomes increasingly integrated into daily life, trust is becoming just as important as capability.

In this episode of Faculty Fridays, Professor James Pang explains why trustworthy AI is important and discusses why AI systems must be designed to be safe, secure, transparent and aligned with human values.

He also shares how his research helps users protect sensitive information when interacting with large language models by restructuring queries before they are processed.

The conversation highlights why trust is becoming a critical consideration as AI plays a larger role in the decisions that affect individuals and organisations.