AIbaseUpdated

Ambi, an AI assistant that listens around the clock, raises $26 million: a former Plaud executive wants to take intelligence out of the…

The AI industry believes that 'the better it knows you, the better it can help you,' at the cost of apps and devices constantly listening, watching, and recording. Though privacy risks are widely known, startups are…

The AI industry has long subscribed to one logic: the better a model knows you, the better it can help you—and the price of 'knowing you' is that apps and devices must listen to you, watch you, and record you nonstop. The privacy risks behind this are common knowledge, yet startups keep charging ahead, betting that the personalized assistant market is big enough to swallow those concerns in one gulp. What former Plaud executive Peter Mo is betting on is precisely a new player on this path: Ambi, an AI app that delivers real-time insights by 'keeping its ears open' and listening to you at all times.

On Monday, Ambi announced the completion of a $26 million funding round, led by Luminous Ventures—an institution that previously invested in Meituan and Unitree Robotics—with original design manufacturer Huaqin also joining in. Ambi simultaneously launched apps for iPhone, Mac, and Apple Watch. They are essentially like meeting recorders, but with one extra switch: users can choose to let them listen all day, capturing context from life and work all along the way.

The traditional approach of meeting recording tools is that, after a call ends, they use a pile of software connectors to automatically run follow-up tasks based on recorded summaries and action items; Ambi wants to go one step further—having its AI agents start working while the meeting is still in progress. The company's promise is that its agents can proactively pop up suggestions or push relevant knowledge in front of you during a meeting, based on your calendar, past meetings, knowledge base, and web data; you can also start something on one device and pick it up on another. When explaining his motivation, Mo put it bluntly: interaction between people and intelligence is too limited—intelligence is often imprisoned inside screens, and with hardware like Plaud's, you have to wait for the recording to reach your phone, upload to the cloud, and finish processing before you get results. 'That's why we built Ambi, so AI can be used anytime, anywhere.'

Ambi's ambitions don't stop at listening. An upcoming desktop app will glance at what you do on your computer, storing the important parts in its 'memory' to give more personalized suggestions—a path that Littlebird, Minimi, and others have already taken. The company is also planning its own hardware: a device with a screen, speaker, camera, and microphone that can both capture context and display information to you.

But 'always listening' naturally steps on the minefield of privacy. Many AI devices today lack a prominent indicator light, so people nearby have no idea they are being recorded; it remains unclear whether Ambi's app and future devices will give third parties sufficiently clear notice. Some meeting-recording products have already landed in lawsuits for recording conversations and data without consent, and Apple's newly launched Audio Intelligence continuous transcription feature has likewise pushed 'technology that is always listening' into the spotlight of public opinion. Mo's response is that ambient data capture can be turned off at any time and is off by default, the company retains no recordings, and the data is encrypted—though he did not specify what kind of encryption is used. The crucial responsibility for disclosure still falls on the user, and he frankly called this an industry-wide problem: 'We have to remind users that they must tell others they are doing ambient capture. As an industry, we haven't solved the disclosure problem. We've put reminders everywhere we can, but if users deliberately hide it, we can't solve that either.'

Ambi currently has offices in four cities—San Francisco, Beijing, Shanghai, and Shenzhen—with a team of more than 50 people, and is hiring for marketing as well as AI engineers to work on speech recognition models. When an AI wants both to get into your ears and to live in your devices, whether it can walk the tightrope between 'being thoughtful' and 'not crossing the line' is probably more worth watching than the funding figure.

Original source

AIbase

Content notes

Original publication and rights belong to the source.

Machine translation · Refer to the original