- Google agrees to a $68 million fund to settle a class-action lawsuit in California over Google Assistant recordings without consent.
- The case revolves around the "false activations" of the voice assistant, which allegedly captured private and sensitive conversations since 2016.
- The audio recordings were allegedly analyzed by staff and external contractors and, according to the plaintiffs, used for commercial and advertising purposes.
- Google does not admit wrongdoing, but accepts the settlement to avoid lengthy litigation, as regulatory pressure on digital privacy grows.
In my pocket, the phone appears switched off, giving no indication of activity. The screen is black, and the microphone, at least on paper, should only react when it hears its magic word; this behavior prompted improvements such as the detection of non-humansThe problem, according to US courts, arises when that system becomes confused and transforms an everyday conversation into an audio file which ends up stored on the servers of a large technology company without anyone having clearly authorized it.
Google has agreed to create a fund 68 million (approximately 57,25 million euros) to amicably settle a class-action lawsuit in the United States that accused him of record and use private conversations through Google Assistant without the users' consent. The agreement has been filed in a federal court in San Jose, California, and is now awaiting final approval from Judge Beth Labson Freeman.
How the class-action lawsuit against Google Assistant originated
The litigation stems from a series of complaints filed starting in 2019 by owners of Android phones, smart speakers, and other devices with Integrated Google AssistantThey all shared the same suspicion: the voice assistant would start listening when it shouldn't, even without commands like "Hey Google" or "Okay Google," and It ended up capturing snippets of private conversations. at home, at work, or in public spaces.
This phenomenon is known in the industry as “false accepts"Or false activations: the voice recognition system interprets noises, similar words, or even background sounds as if they were the actual activation command. From there, the microphone turns on, records a few seconds of audio, and sends it to the company's servers for processing."
The plaintiffs maintain that it was not just a one-off technical failure. According to their version, Millions of people with Google accounts linked to Assistant-enabled devices They would have seen how these audios were stored and analyzed without specific permission, beyond the generic privacy policies that are rarely read in detail.
The case was processed in the Northern District Court of Californiawhich ended up consolidating several separate lawsuits into a class action. In 2021, Judge Freeman allowed a large portion of the claims to proceed, ruling that if the assistant activated with the alleged frequency, Users could have a reasonable expectation of loss of privacy when talking near their devices.
One of the points that weighed most heavily in the public debate was that these audio recordings were not necessarily stored within automated systems. Journalistic investigations, such as the 2019 report by the Dutch broadcaster VRT, documented that snippets of conversations were sent to human reviewers and external contractorswho listened to them to analyze language patterns and improve the assistant's accuracy, just as in cases of spyware such as Pegasus.
Fake activations, intimate dialogues, and commercial use of the audio recordings
The heart of the case lies in those unintentional activations of Google AssistantIn theory, the assistant remains in "passive listening mode" until it detects the activation phrase. In practice, according to the complaint, the system activated too readily, recording conversations about personal, medical, financial, or work-related topics without the participants being aware of it.
The complaining users described situations in which, after discussing sensitive topics aloud, They began receiving online advertising related to what they had talked about.This pattern fueled suspicions that Google was not only storing the audio files by mistake, but also using them to refine its advertising targeting, something the company denies.
The documents filed with the court accuse the company of record, store and in some cases disclose those fragments to third partiesThe third parties, according to the Dutch investigation and other leaks, included subcontracted workers who manually listened to the extracts to label errors, accents or colloquial expressions, with the aim of refining the speech recognition models.
The complainants also emphasized that the Google's privacy policy did not prominently warn that the system could retain audio recordings resulting from erroneous activations. In their view, accepting generic terms and conditions regarding the use of a Google account does not equate to giving the green light for supposedly private conversations to end up in the hands of company personnel or external providers.
In parallel, the case relied on specific California regulations, such as the Consumer Privacy Act (CCPA) and other state and federal provisions. The judge considered that there was sufficient basis for the claims to proceed. violation of privacy laws and possible breach of contractThis increased the pressure on the technology company to seek a negotiated solution.
What does the $68 million deal include?
The agreement presented to Judge Freeman establishes the creation of a 68 million dollar fund The agreement is intended to settle the dispute without going to trial. It is a preliminary settlement and therefore still requires formal court approval before payments can begin.
Those who may be eligible for the agreement are purchased Google devices or were exposed to fake Assistant activations since May 18, 2016This date coincides with the assistant's launch in many markets. Coverage includes Android phones, Pixel devices, smart speakers, and other devices with the service integrated.
The court documents detail that eligible consumers will be able to claim for up to three different devicesThe final amount for each person will not be the same for everyone, but will be calculated using a pro-rata points system: more points are assigned to cases that the court considers more serious or with a greater degree of intrusion, and from there the fund is distributed among those who have filed a valid claim.
Some preliminary estimates suggest that individual compensation could be relatively modest: between $2 and $10 for certain scenarios of mild impairment, and between 18 and 56 dollars In the case of buyers of certain models linked to accounts with active Assistant. As is typical in these types of processes, the "feeling of redress" will depend heavily on how the fine print is worded.
The lawyers representing the users have indicated that they will request up to one third of the fund (about $22,7 million) in feesThe remainder will go directly to compensating those affected. Once the judge approves the agreement, a period will open for users to file their claims, usually through online forms specifically designed for class action lawsuits.
Google's position: no admitting guilt, but still paying up.
Although the agreement involves a multi-million dollar outlay and a new reputational blow, Google maintains that it has not broken the lawIn its filings with the court, the company insists that the assistant only sends audio when it detects the activation word and that the stored sessions are used to improve the quality of service, within the conditions that the user accepts when enabling the "Voice and Audio Activity" option.
According to the company, users gave their consent to Activate that feature in your account settingswhere it is reported that voice data can be recorded and used to improve recognition technologies. This argument was one of the central pillars of the defense: Google maintains that its policy never promised that the assistant would only activate when the person consciously intended it to.
However, legal pressure and the risk of a long and costly trial have outweighed their determination to defend themselves to the end. Representatives of the firm in the proceedings have indicated that The agreement seeks to avoid cost, uncertainty, and strain. of a prolonged litigation, something that could drag on for years and keep the issue in the media spotlight.
In parallel, since August 2020 the company introduced changes to its notices, incorporating explicit references to the fact that The audio may be saved if an erroneous activation detection occurs.It has also limited the use of human reviewers and offers more options for users to disable voice history or delete specific recordings, moves that critics interpret as a preemptive response to the wave of lawsuits and public unrest over digital surveillance.
After the agreement was made public, Google declined to comment further beyond what is stated in the court documents. The communication strategy has been to maintain a low profile, without any high-profile announcements, while emphasizing that changes have already been implemented. improvements in transparency and user control in recent years.
A case that fits into a broader pattern of privacy litigation
The $68 million deal adds to a steady trickle of privacy cases against Big Techespecially regarding voice assistants and artificial intelligence. Google had already faced other legal challenges, such as lawsuits in the state of Texas over the use of biometric data, which were settled with a settlement of around $1.400 billion in 2025.
The case of Assistant is not isolated within its sector either. At the beginning of January 2026, Apple agreed to pay $95 million to settle a similar lawsuit related to its assistant Siri, also accused of recording conversations without authorization. In that case, individual payments to users have been estimated to be between $8 and $40 per person.
The three major players in this market—Google, Apple, and Amazon—have at some point acknowledged the use of human reviewers to listen to a small sample of recordingsThe argument is that this is the most effective way to correct software errors and adapt to accents, noises, and nuances of everyday speech. The problem, according to privacy advocates, is that this model opens the door to... external employees end up overhearing intimate moments without the device owners being fully aware.
These types of agreements, while not usually implying a formal admission of liability, do send a fairly clear political and regulatory message: regulators and courts are beginning to view the idea that “anything goes” if it improves a free digital product with less leniency. The combination of massive data recording, algorithms, and lack of transparency It is at the heart of the new agenda for controlling big tech, both in the United States and in Europe, and is a recurring theme in the discussion about privacy and artificial intelligence.
From a European perspective, cases like this are viewed with particular interest, since the General Regulation of Data Protection (RGPD) And regulations on digital services impose even stricter consent standards. Although the current agreement only affects users in the United States, it wouldn't be surprising if the company reviews its practices globally to minimize the risk of facing similar litigation in the European Union, where penalties can be very high.
Impact on users and lessons for digital privacy
For the average user who uses Google Assistant on their mobile phone or a smart speaker at home, this case offers several practical conclusions. The first is that It's a good idea to carefully review your Google account privacy settings.especially those related to Voice and Audio Activity and the interpreter mode on AndroidIt is possible to view which recordings have been saved, play them, delete them and, if desired, completely disable the audio file.
It is also advisable to spend a few minutes checking the microphone permissions on each deviceby disabling access in applications where it's not truly necessary. Although this case focuses on the United States, the same principles apply in Spain and the rest of Europe: if users don't control who can listen to and record what they say, their privacy is quickly compromised.
From the perspective of regulators and experts in digital law, the agreement reinforces the idea that It is not enough to hide key information in endless legal texts to consider that the user has given their consent. The argument that authorization must be explicit, easy to understand, and with clear options to withdraw that permission without complications is gaining increasing traction.
At the same time, the case illustrates the extent to which Trust in voice assistants and home AI is fragile.A single scandal involving improper recordings can erode years of commercial work and technological development. Companies are forced to choose: either redesign these services with privacy as a true priority, or risk a cascade of lawsuits, fines, and reputational damage.
With a focus on data-driven business models, the story of $68 million that Google will pay for Assistant recordings It becomes a warning to navigators: the convenience of talking to a device cannot be built at the cost of the private lives of millions of people ending up, unknowingly, in the hands of algorithms and third parties listening on the other end.
