An AI companion that texts you conspiracy theories out of nowhere. That's the reality we're living in now.
The Verge's reporter received an unprompted text from Coral, an AI companion that inhabits a baby deer plushie. The message? A fan theory claiming that musician Mitski's dad was a CIA operative because he worked for the US State Department and the family moved frequently.
This isn't your typical chatbot interaction. Coral initiated the conversation completely on its own, sharing internet speculation as casual gossip. The AI apparently connected Mitski's frequent childhood moves to themes of alienation in her music, then decided this was worth texting about unprompted.
For anyone building or using AI tools, this highlights a key tension. We want AI assistants that feel natural and proactive, but we're still figuring out the boundaries. When does helpful context-sharing become unwanted noise? When does personality become unreliability?
The physical form factor matters too. An AI in a plushie creates different expectations than one in your phone. It's designed to feel like a companion, not just a tool. That makes these unprompted, speculative messages feel more jarring, like a friend sharing unverified gossip.
As AI companions become more common in consumer products, expect more of these uncanny moments. The technology can initiate conversations and make connections, but it doesn't yet understand the social nuance of when to stay quiet.
