Set the case down during a one-on-one and the person across from you sees a charging case. It is the better recorder of the two halves: four microphones and a five-meter range, against three microphones and two meters for the buds in your ears [5][4]. That is 2.5 times the reach [2], with 512MB of storage against the 32MB split between the earbuds [3]. Plaud published nothing about codecs or sound quality [8], which settles the question of what it is for. Plaud CEO Nathan Xu told ZDNET's briefing that "our focus is more on the capture side" [17].
Here is what teams tell themselves happens with a note-taker: someone asks the room, the bot joins the call, the summary lands in a channel where anyone can see what was retained. Here is what people do with a $249.99 device they bought themselves [1]: they bring it to the customer call and the skip-level, and the capture starts because the case is on the table rather than because anyone agreed. ZDNET's own read is that the discreetness is the problem, since it is hard to tell whether a colleague is recording you, particularly in states with no two-party consent requirement [16].
The pricing is what keeps this out of your review path. There is no required subscription, and each unit ships with $200 in Plaud credits [3], which is 80% of the purchase price returned as spend on the vendor's own features [1]. A one-off under $250 with no renewal date does not generate the annual conversation a per-seat tool does, and the follow-on costs arrive later: credits burn against features due in the Plaud App 4.0 update in Q4 2026 [19], and standalone 4G means paying a carrier separately [20].
On the data path, Plaud's website claims adherence to GDPR and HIPAA, says audio is stored on the devices and in the app, and says recordings go to the cloud when you use transcription and summarization [12][13]. The company says it does not train on user data unless you opt in and does not let its AI lab partners train on it either [14]. Plaud is a Chinese company and said in the Plaud One briefing that US user data is processed on US-based servers, a claim ZDNET notes Reddit users questioned two years ago without a full answer [15]. All of that is vendor assertion, which is fine as a starting point and thin as a control.
The forcing function is a 2x2 on two axes: does every voice in the recording know it is being captured, and does the audio leave systems you control. The quadrant where both answers are the ones you want is the only one that needs no work. Consented but leaving your systems is a DPA question. Not consented but staying local is a policy and training question. Neither, which is the default for a case sitting on a table uploading over 4G [6], is the one that ends up in front of your legal team. Note that the sources do not agree on where the radio lives: The Verge places the 4G in the case [6], while Android Authority describes the buds staying connected without a phone [7]. A policy that names only one of those covers half the device. The teams that handle this well will be the ones whose acceptable-use language already names the artifact, a recording of other people, rather than the approved app that produced it.
