A three-pound surveillance camera does not contain 23 pounds of copper. Google’s AI said otherwise, and people believed it.

Until recently, anyone typing “how much gold does a Flock camera have” into Google would get a confident, authoritative answer from the search giant’s Gemini AI Overviews. Each Flock Safety camera, the AI declared, “contains about 1 to 5 grams of gold used in its internal circuit boards and wiring.” That would put the gold value of a single camera somewhere between $140 and $700.

The AI also helpfully noted that the cameras contain two to 23 pounds of copper. A camera that weighs three pounds total. The math alone should have killed the claim before it ever appeared in a search result.

It didn’t.

Google’s AI pulled its information from a random Substack account and an obviously fake, AI-generated Instagram post. Both originated in online privacy communities where people opposed to Flock Safety’s AI-powered license plate readers were joking, half-seriously, about precious metals as yet another reason to tear the cameras down. Gemini swallowed the satire whole and served it up as fact.

Flock Safety cameras do contain gold and copper. Every piece of consumer electronics does. The trace amounts found on circuit boards and in wiring are standard across the industry, and there is nothing unusual or valuable about the materials inside a Flock camera compared to what sits inside a cheap router or a Ring doorbell.

But the AI Overview didn’t bother with that context. It cited its sources with the same deadpan confidence it uses to summarize medical research or explain tax law. No flags, no caveats, no indication that the underlying material was a joke post on social media.

Flock cameras have already become frequent targets of vandalism. The company’s surveillance technology, which uses automated license plate recognition to track vehicle movements, has drawn fierce opposition from privacy advocates and civil liberties groups. A Florida sheriff recently yanked his county’s Flock units over concerns about how a single detective was using the data.

Adding a financial incentive to an already volatile situation is reckless, even if the incentive is completely fabricated. Someone searching for information about these cameras and seeing an authoritative Google result claiming they contain hundreds of dollars in gold might not stop to question why a three-pound device supposedly holds 23 pounds of copper.

Google has since corrected the AI Overview, but the episode exposes something deeper about the architecture of modern search. Gemini does not verify. It aggregates. It treats a throwaway Substack post and a fake Instagram graphic with the same weight it gives peer-reviewed research, because it cannot tell the difference.

The system is designed to sound certain, not to be correct. This is the same AI infrastructure that Google wants to place at the center of how billions of people find information. The same technology that automakers are rushing to embed in their dashboards and driver-assistance systems.

A joke about gold in a surveillance camera traveled from a privacy forum to a fake Instagram post to the top of Google search results, where it sat unchallenged, looking every bit as reliable as the weather forecast.