Ambi's all-day listening apps leave disclosure to whoever is using them

In the example Mistlabs gives, a user sits through a meeting about competitors while several agents fact-check what is being said ^[8]. Before anyone leaves the room, Ambi may have compared the rival products and drafted a proposal on the user's MacBook ^[8].

That is a pitch to the person who spends the day in meetings, and the same person carries the duty to tell the room. Peter Mo, Ambi's co-founder and CEO, told TechCrunch, "We have to inform our users that they have to tell others about ambient capture." ^[15] He called disclosure an industry-wide problem ^[18]. "We have tried to put the information about disclosure in every place possible, but if users try to hide the capture, we can't just solve the problem," he said ^[16].

The easy assumption is that a user who switches on all-day capture will tell everyone nearby. Mo's own words allow for the user who hides it, and the company says it cannot fix that ^[16]. Neither report includes usage figures, such as how many people turn ambient capture on or how long they keep it on.

TechCrunch lists the field. Littlebird and Minimi already do something similar to Ambi's upcoming desktop app, which will look at the work users do and store important parts in its memory ^[19][13]. Ambi's planned device adds a camera to the microphones, along with a screen and a speaker ^[4]. According to TechCrunch, startups are building such apps anyway, on the belief that the market for personalized assistance is big enough to swallow the privacy concerns ^[17].

The startup said it stores no recordings and encrypts the data, but it did not specify what kind of encryption ^[6]. The sources describe other things it keeps. SiliconANGLE says the software's read on how a person likes to work sharpens over time ^[7], and TechCrunch says the desktop app will hold important parts of a user's work in memory ^[13].

We would withhold approval of ambient mode on work devices until the company states what encryption it uses and what the learned profile holds. The tradeoff is losing the mid-meeting suggestions that set Ambi apart from an ordinary note-taker ^[3].

Two tests sort any tool of this kind: whether it hears people other than the user, and whether it keeps anything after the call, whether recordings or a learned profile. A tool that hears only the user and keeps nothing is a dictation tool and needs little review. One that hears only the user but keeps a profile needs a retention answer. One that hears others and keeps nothing needs a disclosure rule. One that does both needs both. Ambi falls in the last group on the sources' description, since it learns from the conversations it hears and its desktop app is to store parts of what users work on ^[7][13].