Electric Plant, a startup, says it has built an AI-powered translator that reads the faint electricity running through plant stems and turns it into something humans can interpret.
That claim is the entire reported news. Nothing public so far describes what data the model learned from, how it was tested, what the product looks like, or whether anyone can use it. Even "translator" is the company's own framing, and nothing reported says what the output actually is.
The premise does rest on real biology. Plants carry electrical signals, and the Venus flytrap is the textbook case: it snaps shut after its trigger hairs are touched. What remains unshown is the jump from "plants produce electrical signals" to "plants are saying something" that software can decode. A flytrap's signal is a reaction to a touch. Calling a stream of voltage readings a language is a much bigger claim.
That gap is where this kind of project lives or dies. A model trained on noisy sensor data can find patterns in almost anything, so the pattern alone proves little. The more useful question is whether the model can match a signal to an event researchers already know about, such as a plant being cut, watered or shaded, and do it on plants it has never seen.
AI projects that try to decode animal sounds face the same problem of checking a translation against something independent. Until Electric Plant shows how it handles that check, its translator is a pitch with an interesting biological hook, not a result.