Ex-Google Engineer Charged With Selling Trade Secrets To Uber

Former Google engineer and self-driving pioneer Anthony Levandowski is being charged with stealing closely guarded secrets and selling them to Uber, The Wall Street Journal reported on Tuesday (Aug. 27).

Levandowski was charged with 33 counts of trade secret theft in an indictment filed by the U.S. Attorney’s office in San Jose, California. He could be sentenced to up to 10 years and fined $250,000 per count, $8.25 million altogether.

Levandowski was accused of stealing years of top-secret information from Google, which prosecutors called the crown jewels of the company. That included Google’s breakthroughs in lidar, a key piece of technology that enables self-driving cars to detect what’s around them.

The lawsuit stems from a 2017 accusation filed by Waymo, a self-driving car pioneer spun off from Google. Although Uber agreed to pay Waymo $245 million in a 2017 civil suit, the judge said a criminal investigation would be opened.

Although Levandowski wasn’t a defendant in that case, the allegations were referred to prosecutors. San Francisco U.S. District Judge William Alsup said the record contained “ample evidence” of a breach of confidence.

In a statement reported by WFTV, Levandowski’s attorneys maintained his innocence. Prosecutors say the probe is ongoing, but would not clarify whether Uber and its founding CEO Travis Kalanick are targeted. Prosecutors say Google and Uber cooperated in the investigation.

“He didn’t steal anything from anyone,” the statement said. “This case rehashes claims already discredited in a civil case that settled more than a year and a half ago.”

The self-driving and robotaxi company Waymo spun off from Google in 2016 and recently teamed with the AI firm DeepMind to better train self-driving software in a way that mimics Darwinian evolution.

Waymo and DeepMind are owned by Alphabet, which is also Google’s parent company. The companies use a method called population-based training (PBT), which lessens false positives when the software performs actions like placing boxes over the moving objects it sees in its sensors. The new training method also uses 50 percent less time and resources than previous methods.