Google in public comments to the NTIA ahead of its report it said it expects “to see greater attempts to alter, degrade, deceive and steal” models. But he added that his secrets are guarded by a “security and reliability organization made up of engineers and researchers with world-class experience” and that he was working on “a framework” that would involve a committee of experts to help regulate access to Models and their weights.
Like Google, OpenAI said in the comments. to NTIA that open and closed models were needed, depending on the circumstances. OpenAI, which develops models like GPT-4 and services and applications that rely on them, like ChatGPT, formed its own security committee on its board of directors last week and this week published details on his blog about the security of the technology he uses to train models. The blog post expressed hope that transparency will inspire other laboratories to adopt protective measures. He did not specify from whom the secrets were to be protected.
Speaking alongside Rice at Stanford, RAND CEO Jason Matheny echoed her concerns about security breaches. By using export controls to limit China’s access to powerful computer chips, the United States has hindered the ability of Chinese developers to develop their own models, Matheny said. He claimed that increased his need to steal AI software directly.
By Matheny’s estimate, spending a few million dollars on a cyberattack that steals pesos from AI models that cost an American company hundreds of billions of dollars to create is well worth it for China. “It’s really difficult and it’s really important, and we’re not investing enough nationally to do it right,” Matheny said.
The Chinese embassy in Washington, D.C., did not immediately respond to WIRED’s request for comment on the theft allegations, but in the past has described such claims as baseless smears by Western officials.
Google has said it notified authorities about the incident That became the US case of stealing AI chip secrets for China. While the company has described maintaining strict safeguards to prevent the theft of its proprietary data, court documents show that it took Google considerable time to catch the defendant, Linwei Ding, a Chinese national who has pleaded not guilty to federal charges.
The engineer, who also goes by the name León, was hired in 2019 to work on software for Google’s supercomputing data centers, according to prosecutors. For about a year starting in 2022, he allegedly copied more than 500 files with sensitive information to his personal Google account. The scheme worked in part, court documents say, when the employee pasted information into Apple’s Notes app on his company laptop, converted the files to PDFs and uploaded them elsewhere while evading Google technology intended to detect that type of exfiltration.
While participating in the alleged theft, the United States claims that the employee was in contact with the CEO of an artificial intelligence startup in China and had moved to start his own Chinese artificial intelligence company. If he is convicted, he faces up to 10 years in prison.
