Algorithmic Hallucination: How AI Turned a Joke into a 'Fact'
In August 2026, an incident occurred that vividly illustrates the risks of uncontrolled information dissemination by generative models. The AI Overview algorithm in Google's search engine began providing users with false data regarding the traffic surveillance cameras of the company Flock Safety. The artificial intelligence confidently claimed that each autonomous license plate recognition camera contained between 1 and 5 grams of gold, as well as between 0.9 and 10.4 kilograms of copper. Based on this data, the value of the gold alone in a single device was estimated at approximately $650.
The critical error lay in the fact that the AI failed to perform an elementary check of the physics of the process: the stated mass of copper, weighing several kilograms, far exceeded the mass of the device itself, which is only 1.36 kg. Despite this obvious contradiction, the algorithm continued to disseminate the information, leading to real-world consequences.
Wave of Vandalism and Context of Distrust
The spread of false information about "golden cameras" did not go unnoticed. Users who believed in the possibility of easy profit began committing acts of vandalism against Flock Safety equipment. The story began against a backdrop of growing scandals surrounding the surveillance system. Previously, there were reports of abuses where individual law enforcement officers used the system for personal surveillance. Public dissatisfaction with constant monitoring created fertile ground for a hoax encouraging the destruction of cameras.
Although people have spread false information before (for example, in 2020, cell towers were burned down without AI assistance), in this case, the algorithm played the role of a catalyst, giving an absurd theory the appearance of a technical fact.
Reality vs. Fiction: What's Inside the Camera
After the story gained widespread resonance, Google was forced to adjust the operation of its AI Overview. Updated search results now contain a warning: stories about grams of extractable gold and kilograms of copper are someone's joke and the result of neural network hallucinations. The actual content of precious metals in such devices is comparable to what is typical for any compact electronics.
Gold is indeed present in the cameras, but only in trace amounts — mainly in the form of microscopic coatings on contacts, connectors, and individual elements of printed circuit boards. The same applies to copper: it is used in conductors and board traces, but its quantity is negligible and measured in micrograms, not kilograms.
Contradictory Data
There is a discrepancy in estimates of the scale of the consequences. On the one hand, sources confirm that the Google algorithm did indeed provide false data about the content of gold and copper, which provoked vandalism. On the other hand, the exact number of damaged cameras and the total material damage caused to equipment owners as a result of this "gold rush" remain unknown. The story omits specific figures of destroyed property, focusing on the fact of the AI error itself.
Algorithmic Responsibility in the Era of Generative Content
The story of the "golden cameras" became another characteristic example of generative AI repeating and disseminating unreliable facts. The platform failed to perform an elementary check of the plausibility of the data, which led to real damage. This case raises the question of who should be held responsible for the actions of algorithms: developers, users, or the machines themselves. While humans are well-versed in spreading fake news without AI prompts, the scale and speed of neural networks make such errors particularly dangerous.
