Google Is Telling People Flock Cameras Are Full Of Gold

Google’s Gemini AI Overviews feature has been telling users that Flock Safety’s surveillance cameras are packed with valuable gold and copper, a claim that appears to have originated from fabricated reports circulating online. It isn’t true. The cameras do contain both metals, but only in trace amounts consistent with any consumer electronics device pulled off a shelf, as Futurism reported.

Flock’s AI-powered license plate readers have grown more common and more controversial. As a result, vandals have increasingly targeted them. While privacy concerns drive much of that activity, Google’s AI may have added a financial incentive built on false information.

Until recently, searching “how much gold does a Flock camera have” on Google returned an AI Overview stating with confidence that each camera “contains about 1 to 5 grams of gold used in its internal circuit boards and wiring.” If accurate, that would place between $140 and roughly $700 of gold inside every unit, a figure that would make theft far more lucrative than the hardware itself.

The AI Overview went further. It claimed the cameras can hold two to 23 pounds of copper. As Futurism noted, that assertion collapses under basic scrutiny, since a Flock camera weighs just three pounds. A single unit cannot contain 23 pounds of anything.

The source of the claims traces back to online privacy communities, where users half-seriously floated the idea of precious metals inside the cameras as one more reason to tear them down. Google’s AI then treated those jokes as fact. The system cited a random Substack account and an obviously AI-generated Instagram post, both of which described the supposed gold and copper content in detail. Neither qualifies as a credible technical source. Yet the AI Overview presented the numbers without qualification.

The episode illustrates a persistent weakness in generative AI tools used for research. They can return accurate information. But they can also surface fabricated figures while citing unverified or deliberately misleading sources. In this case, the model pulled from material that was either satirical or entirely synthetic, and then repackaged it as authoritative fact for anyone running a straightforward search query.

Anyone acting on the AI’s figures would find no meaningful payout. None. Stripping a camera for its metals would yield only the negligible traces found in comparable electronics, not the multi-hundred-dollar haul the AI Overview described.