Google And Meta Shipped AI Tools That Fake Evidence—Then Pulled Them Days Later
For one day in late July, anyone could type a sentence into Google Earth and get a fake satellite image rendered on top of Google's real imagery of a real place. Researchers made a nuclear plant in Iran and a bomb crater at a Gaza hospital within hours. Google pulled the feature the next day.
For one day in late July, anyone could type a sentence into Google Earth and get a fake satellite image rendered on top of Google's real imagery of a real place. Researchers made a nuclear plant in Iran and a bomb crater at a Gaza hospital within hours. Google pulled the feature the next day.
In this video: why the rollback restores the feature list and not trust. Meta ran the same play three weeks earlier with Instagram photos. The watermark that was supposed to flag the fakes came up empty when Google's own chatbot checked for it. And every demo prompt, add a cabin to this lakefront lot, is one noun swap from a fraud prompt, add storm damage to this roof.
I'm Lars Daniel, digital forensics expert and Forbes contributor. I examine phones, computers and digital files, and testify about what I find in state and federal courts.
This video is general information, not legal advice.
