Close Menu
clearpathinsight.org
  • AI Studies
  • AI in Biz
  • AI in Tech
  • AI in Health
  • Supply AI
    • Smart Chain
    • Track AI
    • Chain Risk
  • More
    • AI Logistics
    • AI Updates
    • AI Startups

UAE startup Tandia attracts VC investment in AI-driven data monetization product

February 23, 2026

Trust ANC obtains patent for “AI Building” automation system

February 23, 2026

SK Networks invests more in AI startup Upstage

February 23, 2026
Facebook X (Twitter) Instagram
Facebook X (Twitter) Instagram
clearpathinsight.org
Subscribe
  • AI Studies
  • AI in Biz
  • AI in Tech
  • AI in Health
  • Supply AI
    • Smart Chain
    • Track AI
    • Chain Risk
  • More
    • AI Logistics
    • AI Updates
    • AI Startups
clearpathinsight.org
Home»AI Research Updates»UF researchers develop new training method to help AI tools learn safely
AI Research Updates

UF researchers develop new training method to help AI tools learn safely

December 24, 2025003 Mins Read
Share Facebook Twitter Pinterest Copy Link LinkedIn Tumblr Email Telegram WhatsApp
Follow Us
Google News Flipboard
Adobestock 432173677 fotor 20251217111116.jpg
Share
Facebook Twitter LinkedIn Pinterest Email Copy Link

As artificial intelligence becomes integrated into everyday life, UF researchers are working to ensure the technology learns safely. A new paper from the University of Florida and Visa Research presents a training method designed to prevent AI models from memorizing sensitive information – a growing privacy risk in modern machine learning systems.

The work, entitled “Deep Learning with Plausible Deniability”, was presented in early December at NeurIPS 2025one of the most prestigious AI conferences in the world. The paper is led by UF Ph.D. student Wenxuan Bao and UF associate professor Vincent Bindschaedler, Ph.D., in collaboration with Visa Research.

“We don’t want to design systems that might be the smartest, without taking into account how they handle sensitive data,” said Bindschaedler, based in UF Department of Computer, Information Sciences and Engineering. His work focuses on creating what he calls “trustworthy machine learning,” an area that includes privacy, security, and interpretability.

Why AI memorization is important

During training, AI models repeatedly analyze their data sets to improve performance. But sometimes they focus on specific details, like a phone number, medical records, or even a person’s name, instead of general patterns.

“It basically remembers very specific detailed information from its training corpus,” Bindschaedler said. “We actually have techniques to probe the AI ​​to try to recover it.”

Since many models are publicly available, memorized data can potentially be extracted.

“If a widely used system allowed a person’s medical record to be stored and someone knew that that was the case, they could essentially pull it out,” Bindschaedler said. “That would be pretty catastrophic.”

A simple privacy test with big impact

The team’s new technique adds a quick “privacy check” during training. If an update could reveal information about an individual data point, the model simply wouldn’t use it.

“It’s a very simple natural thing,” Bindschaedler said. “We train the AI ​​in the normal way, but we just have this kind of extra control that says, ‘Hey, could this potentially leak information?’ And if that’s the case, then you’re not updating.

This check is based on the idea of ​​plausible deniability: the model must never perform an update that can be linked to a single record; it must be explainable by several different subsets of data. Otherwise, the update is rejected.

“It’s just going to start again,” he said. “He just throws it away and then moves on to the next batch.”

Increasing the visibility of UF in AI research

For Bindschaedler, the acceptance of NeurIPS signals both the scientific impact and growing presence of UF in the AI ​​landscape.

“NeurIPS is truly the flagship place for academic AI research…where work like this will get the most visibility and the most impact,” he said.

He added that publishing in top venues is part of advancing UF’s strategic focus on AI.

“Given UF’s goal of advancing AI … it makes sense to emphasize these types of most visible flagship locations,” he said.

The team is currently exploring how the method can be extended to new applications and strengthened theoretically.

“This work introduced a new technique – there’s a lot we don’t know about it,” Bindschaedler said. “We think there are some interesting applications…that could be of practical use.” »

Follow on Google News Follow on Flipboard
Share. Facebook Twitter Pinterest LinkedIn Tumblr Email Copy Link

Related Posts

ChatGPT’s in-depth research tool adds a built-in document viewer so you can read its reports

February 22, 2026

Google AI January announcements

February 21, 2026

Announcing our latest Gemini AI model

February 21, 2026
Add A Comment
Leave A Reply Cancel Reply

Categories
  • AI Applications & Case Studies (70)
  • AI in Business (402)
  • AI in Healthcare (314)
  • AI in Technology (394)
  • AI Logistics (52)
  • AI Research Updates (131)
  • AI Startups & Investments (325)
  • Chain Risk (88)
  • Smart Chain (116)
  • Supply AI (105)
  • Track AI (70)

UAE startup Tandia attracts VC investment in AI-driven data monetization product

February 23, 2026

Trust ANC obtains patent for “AI Building” automation system

February 23, 2026

SK Networks invests more in AI startup Upstage

February 23, 2026

AI in healthcare has evolved faster than expected

February 23, 2026

Subscribe to Updates

Get the latest news from clearpathinsight.

Topics
  • AI Applications & Case Studies (70)
  • AI in Business (402)
  • AI in Healthcare (314)
  • AI in Technology (394)
  • AI Logistics (52)
  • AI Research Updates (131)
  • AI Startups & Investments (325)
  • Chain Risk (88)
  • Smart Chain (116)
  • Supply AI (105)
  • Track AI (70)
Join us

Subscribe to Updates

Get the latest news from clearpathinsight.

We are social
  • Facebook
  • Twitter
  • Pinterest
  • Instagram
  • YouTube
  • Reddit
  • Telegram
  • WhatsApp
Facebook X (Twitter) Instagram Pinterest
© 2026 Designed by clearpathinsight

Type above and press Enter to search. Press Esc to cancel.