Regular audits would build trust, confidence in AI
Regular audits would build trust, confidence in AI
https://news.cornell.edu/stories/2026/05/regular-audits-would-build-trust-confidence-ai
Publish Date: 2026-05-13 09:15:00
Source Domain: news.cornell.edu
Here are six key points summarizing the article:
– The article discusses the uncertainty surrounding the reliability and risks of artificial intelligence (AI). It draws a parallel between current AI technology and the early stages of indoor plumbing and running water where safety wasn’t immediately assured.
– One solution to build confidence in AI, according to J. Nathan Matias, is conducting regular audits and safety testing, similar to how the water and car industries came to ensure safety over time.
– Matias, assistant professor of communication, and director of the Citizens and Technology Lab, co-authored “Auditing AI,” aiming to teach consumers how to evaluate AI based on its expected performance and its potential to cause harm.
– Matias compares AI auditing to regular car inspections, emphasizing that AI systems’ behavior can change over time, requiring continual monitoring and maintenance to ensure reliability.
– Potential auditors need to understand human and organizational needs, possess technical and statistical knowledge, and be skilled in building trustworthy evidence, integrating technical and communication skills.
– Matias believes striking a balance between useful AI audits and encouraging innovation is crucial. While such evaluations may slow down innovation, they help prevent dangerous products that can harm public trust, ensuring that consumers feel safe and secure in using AI in the long term.
