Lyu urges protections against AI scams in Assembly testimony

UB computer scientist Siwei Lyu delivered a clear message to New York State lawmakers on Tuesday regarding the dangerous intersection of personal information and artificial intelligence.

“Traditional data mining already reveals sensitive information about our lifestyles, beliefs and preferences. The use of this data to train generative AI systems makes the risks even greater,” said Lyu, who appeared before members of the Assembly Standing Committee on Consumer Affairs and Protection, and the Assembly Standing Committee on Science and Technology.

The public hearing is available on the Assembly’s website. Lyu appears roughly 2 hours and 19 minutes into the session.

An expert in digital media forensics, computer vision and machine learning, Lyu said that “when personal images or voice recordings are collected without adequate safeguards, they can be used to create hyper-realistic synthetic media, commonly known as deepfakes, that can easily deceive consumers and users.”

As a result, he added, “we need strong, sensible safeguards to protect New Yorkers and their communities.”

Lyu, a SUNY Distinguished Professor and SUNY Empire Innovation Professor in the Department of Computer Science and Engineering, co-directs the university’s Center for Information Integrity, a multidisciplinary group dedicated to combating the spread of unreliable information.

His research includes creating the DeepFake-O-Meter, an open platform used worldwide by journalists and the public to identify false videos, pictures and other digital content. He was among a handful of scholars, policy experts and others invited to address the two lower house committees.

“We’re convening today because the stakes around consumer protection and data privacy have never been higher,” said Assemblymember Nily Rozic, who chairs the consumer affairs and protection committee.

Assemblymember Steven Otis, who chairs the science and technology committee, previously visited UB with SUNY Chancellor John B. King Jr. and other elected officials to learn about UB’s efforts to promote AI for the public good.

“Data privacy is very much linked to cybersecurity vulnerability, and data privacy is also very much linked to what is happening in the fast-moving world of artificial intelligence,” Otis said.

Lyu offered steps lawmakers can take to help safeguard the public. These include:

• Transparency: Companies should be required to clearly report to the public about what data they collect, how it is used and with whom it is shared, he said.
• Data minimization: Rules should guide data collection practices, ensuring that companies gather only information that is strictly necessary.
• Accountability mechanisms: To prohibit the misuse or mishandling of personal information in AI systems, he said, New York State should create independent AI auditing guidelines and rules.
• Practical defenses: These are tools to reliably detect AI-generated media and methods to remove certain user data from models.
• Public education investment: This is especially critical for vulnerable groups, including teens and older adults, he said.

“New York State has already shown national leadership and forward-thinking vision through the Empire AI initiative,” Lyu said. “But we cannot underestimate the risk posed by unregulated data collection and the misuses of generative AI technologies.”