Claude finds TikTok, Instagram bypassing microphone permissions to listen to you

A viral claim describes using Claude Code to examine an iPhone after TikTok showed the writer’s girlfriend breakup videos shortly after he had discussed ending their relationship near his phone. The writer says the tool detected hundreds of background microphone activations by TikTok, Instagram, Facebook and Snapchat over 90 days, along with locally cached keywords and timestamps. Examples allegedly included “breakup,” “moving out,” “wedding venue,” “protein powder” and “flight to Tokyo,” each linked to nearby conversations. The writer also claims a second phone recorded the same keywords, and says disabling microphone access, background refresh, “Hey Siri” and nearby-interaction permissions stopped the activity. Within two days, both phones supposedly received generic, less personalized feeds.

The account presents these observations as proof that the apps secretly listen and use conversations to shape recommendations, but it supplies no independently verifiable evidence in the text. The accompanying community context questions its credibility: the writer reportedly referred to a wife in a recent claim and a girlfriend here, has a history of sensational AI-related claims that readers allege were fabricated, and promotes AI products. The content is therefore both a serious privacy accusation and a personal account whose stated technical findings and conclusions are disputed.