In 2024, a fraud gang used an AI-generated video of "President Biden's TV speech" to claim that "Tesla will be nationalized", causing its stock price to plummet by 23%, and short sellers to illegally profit $900 million. The video was even briefly mounted through a loophole on the CNN official website, which was extremely confusing.
Core team actions:
CFIT digital forensics expert Carlos Miguel used the "light field analysis method" to find that the frequency of the swinging of the White House curtains in the video did not match the real physical laws; speech synthesis expert Linda Chen identified the original material from Biden's 2019 speech clips through voiceprint comparison.
Partners:
DARPA: Provide "deep fake real-time detection chips" embedded in mainstream video platforms.
Nasdaq: Emergency suspension of trading to cooperate with the investigation of short-selling accounts.
OpenAI: Ban the API keys involved in the case and update the watermark rules for synthetic content.
Legal controversy:
Citing Section 10(b)-5 of the Anti-Market Manipulation Act, it defines "AI falsification of public figures' statements" as securities fraud for the first time, and the principal offender faces 20 years in prison.
Social impact:
The U.S. Securities and Exchange Commission (SEC) requires all listed companies to disclose the risks of "AI-generated major statements", and the case was selected as one of Stanford University's "Top Ten AI Ethics Cases".