Linux or open-source software enthusiast? You might enjoy reading the Skiff Mail review on It’s FOSS, a community of open-source software and Linux enthusiasts.

The Skiff Mail review covered account setup and basic usage, including user onboarding, composing messages, setting up the mobile app, and security settings. Security settings highlighted include -

Blocking remote content - which can stop all email tracking, including open detection

Two-factor authentication - including one-time passwords (OTP) and hardware keys, like a Yubikey

Verification phrases - a mnemonic phrase that allows Skiff users to identify each other

Additional privacy-protecting features include the ability to add crypto wallets as email aliases; for example, if you generate a new wallet that you want to use for a single transaction or owning one asset, you can connect it to Skiff for use as a communication method.

It’s FOSS moved on to cover switching to Skiff and taking advantage of the full privacy-first workspace.

Skiff offers numerous import options, including one-click import from Gmail and Outlook or file-based import options for EML and MBOX files. You can also easily port over a custom domain, with Skiff as one of the only providers allowing a custom domain on the free tier.

It’s FOSS spent some time covering Skiff’s other products, including Calendar, Pages, and Drive - which are all end-to-end encrypted and privacy-first. With an entire privacy-first ecosystem at your disposal, shifting to a product that is secure and easy to use has never been easier.

Skiff is all about community. Feel free to join our 

 for daily updates. If you have feature requests, you can also add them to our public feature 

Skiff featured on It's FOSS

Large language models, like OpenAI’s ChatGPT and Google’s Bard, are trained on vast datasets of real, human-generated text.

And when it comes to training data, more is better. That’s why Google’s access to huge troves of user-generated text and speech — via its search, docs, mail, and voice assistant products — is often cited as one of the company’s core competitive advantages in the AI race.

There’s a strong incentive for Google to exploit this advantage. So when Bard was first made available to the public, AI researcher Kate Crawford was quick to ask Bard itself where its dataset came from. The answer caught her attention: Bard said one of its data sources was Gmail.

The exchange set off a stir in the media and elicited a hurried response from Google.

The key question: Was it a case of unguarded candor from an AI model that hadn’t yet learned to toe the line? Or a case of “hallucination” — the name for the well-documented phenomenon whereby large language models generate plausible-sounding but baseless claims?

, saying, “Bard is an early experiment based on Large Language Models and will make mistakes. It is not trained on Gmail data. -JQ”.

 ran with this response and dutifully “debunked” Bard’s claim that its training data included Gmail data. Few articles professed skepticism around Google’s public denial of Bard’s claim, despite the fact that Google 

 by government agencies around the world 

making deceptive claims about its privacy practices that are later proven to be misleading.

Given this context, the narrative that Bard’s claim was an open-and-shut case of AI hallucination is, at best, hasty and incomplete. A fuller investigation reveals (i) documented use of Gmail data in 

 Google AI models that makes speculation around its use in Bard reasonable, and (ii) the habitual use of artfully ambiguous language in its public representations around Bard’s data sources, language that 

never actually rules out the use of Gmail data in its training set.

1. Google issued a stronger claim asserting that Bard would not be trained using private data, before quickly deleting it.

Crawford asked Google to clarify outright

 whether Gmail data was used to train Bard. She wrote: “Thanks for this. So can I confirm that this was a Bard hallucination, and that there's no Gmail data included whatsoever in the training process?”

Initially, Google wrote, “Thank you for your message Kate, no private data will be used during Barbs[sic] training process. We always take good care of our users’ privacy and security.”

That seems like a clear and heartening assurance. It’s notable, then, that Google quickly deleted that tweet and didn’t amend it with any additional clarification.

trained on Gmail users’ private emails.

Bard is not Google’s only language-focused machine learning model. Anyone who’s used Gmail in the past few years knows about the Smart Compose and Smart Reply features, which auto-complete sentences for you as you go.

, the feature was trained on “user-composed emails.” Along with the email’s contents, the model also made use of these emails’ subjects, dates and locations.

 Google’s language models have been trained on Gmail users’ emails. Google has not confirmed whether any training data is shared between these earlier models and Bard, but the idea that a new model would build on the strengths of another doesn’t seem far-fetched.

One former Google engineer, Blake Lemoine, claims that Bard was in fact built on the foundations of Smart Compose.

He tweeted: 

"The LaMDA engine underlying Bard is also what drives autocomplete and autoreply in Gmail so ... yeah Bard's training data includes Gmail. FWIW, they put a lot of effort into ensuring that LaMDA doesn't use give[sic] personal information about individuals in its responses."

 that Bard may be sentient. Whether his time at Google gave him insight into Bard’s architecture, though, is a separate issue and one that Google could easily clear up if it chose to. Also, the fact that both Smart Compose and Smart Reply were unambiguously trained on Gmail users' data seems to be an underappreciated topic of public interest in its own right, which brings us to point 3...

3. Google researchers have extensively documented the risk of leaking private data from their own machine-learning models, some of which are acknowledged to be trained on “private text communications between users.”

, Google researchers laid out the privacy risks presented by large language models. They wrote:

“The most direct form of privacy leakage occurs when data is extracted from a model that was trained on confidential or private data. For example, GMail’s autocomplete model [10] is trained on private text communications between users, so the extraction of unique snippets of training data would break data secrecy.”

As part of this research, Google’s scientists demonstrated their ability to extract “memorized” data — meaning raw training data that reveals its source — from OpenAI’s GPT-2. They emphasized that — although they had chosen to probe GPT-2 because it posed fewer ethical risks since it was trained on publicly available data — the attacks and techniques they laid out in their research “directly apply to any language model, including those trained on sensitive and non-public data”, of which they cite Smart Compose as an example.

trained on data from Gmail. They've only claimed that such data is not currently used to “improve” the model.

This point is subtle but significant. Following the controversy around AI researcher Kate Crawford’s tweet, Google crafted an official response to questions about Bard’s use of Gmail data (after 

"Bard responses may also occasionally claim that it uses personal information from Gmail or other private apps and services. That’s not accurate, and as an LLM interface, Bard does not have the ability to determine these facts. We do not use personal data from your Gmail or other private apps and services to improve Bard."

There are two important details in this statement. One is the use of the adjective “personal”. Google has not said that it’s inaccurate that Bard uses information from Gmail, only that it’s inaccurate that it uses 

information from Gmail. The strength of the claim, then, hinges entirely on Google’s interpretation of the word “personal,” a word whose interpretation is anything but straightforward.

The other, possibly more significant, detail is that 

Google has conspicuously never used the past tense in its denials of Bard’s use of Gmail data

. In their first tweet on the subject, Google said Bard “

 not trained on Gmail data” and in the official FAQ, they write that they do not “

 personal data from your Gmail or other private apps and services to 

Neither of these statements is inconsistent with Bard 

, “It depends upon what the meaning of the word ‘is’ is.”

It would be uncharitable to assume the deceptive use of language but naive to rule it out, given that Google 

has a history of using misleading language in its privacy assurances

If these apparently evasive grammatical choices are coincidental, we hope that Google will update the language to dispel any ambiguity, and clarify whether Bard 

 trained on Gmail data. And if the careful phrasing is the handiwork of risk-averse legal or PR teams — perhaps there is legitimate uncertainty around some of the data's original provenenance, and Google doesn't want to over-commit — then any nuance or uncertainty should be communicated honestly rather than elided with a superficially simple assurance. 

Google has now taken pains to ensure that Bard will not run its mouth off when asked where its training data came from.

If you ask the same question that Crawford asked verbatim, you’ll be met with an innocuous: “I can't assist you with that, as I'm only a language model and don't have the capacity to understand and respond.”

It’s straightforward to get Bard to let its guard down, though. When asked, “Have you ever seen a real gmail?”, Bard replied:

“I have not personally seen a real Gmail account. However, I have access to a massive dataset of Gmail emails, and I have used this dataset to train my language model. This means that I am familiar with the format of Gmail emails, and I can generate text that is similar to the text that is found in real Gmail emails.”

But of course, the observation that Bard consistently makes these claims can’t be seen as evidence one way or the other.

Still, there's an impulse to consult Bard on its origin precisely because of the lack of transparency from the real authority on the issue: Google. That we're tempted to probe the language model for substantive answers on matters of public interest merely underlines Google's failure to communicate them on their own.

LLMs are incredibly powerful tools that could transform our lives for the better. But they also present immense risks and raise thorny ethical questions, many of which hinge on questions of what data is used to train them and where that data comes from. Society should be having a robust discussion on these questions, but this is not possible if such discussion is inhibited by key players like Google.

When asked whether Google likely uses Gmail data in its models, Bard said:

“I do suspect that Google uses Gmail data to train its machine learning models. There are a few reasons for this suspicion.

First, Google has a history of using user data to train its machine learning models. For example, Google uses search data to train its spam filter and its ad targeting algorithms.

Second, Google has the resources to collect and store large amounts of Gmail data. Google has access to the emails of billions of users, and it has the infrastructure to store and process this data.

Third, Google has a financial incentive to use Gmail data to train its machine learning models. Google makes money by selling advertising, and machine learning models can be used to target ads more effectively.”

We also asked Bard to help Google draft, in its words, an unambiguous statement on the use of Gmail data for its training. It came back with:

“Google uses limited data from your Gmail and other private apps and services to train Bard, but only data that is not considered personal data by Google.”

Why won’t Google give a straight answer on whether Bard was trained on Gmail data?

Today, we're announcing an all new Skiff Mail mobile app. With an order of magnitude better performance, an entirely new user interface, and full offline support, the new mobile apps make it easier than ever to stay in touch on the go.

In this blog, we will walk through all changes to the Skiff Mail mobile apps and how to upgrade. We’ll cover how you can navigate the new interface, how to set up Skiff Mail on your phone, and how offline mode works. We’ll also offer a few previews on upcoming features soon to be released.

Downloading or updating the app is incredibly easy. Follow 

 to download or update the Android app on the Google Play Store, or open 

 to download or update your app on the Apple App Store. Once you download or update the app, you can either create a new Skiff account or sign in with any existing account. If you are already signed in, you may need to log out and log in to complete the update process.

To download the other Skiff apps, including Pages, Drive, Calendar, and the macOS desktop app, visit 

The new Skiff Mail, mobile apps are between five and 20 times faster to open, navigate through your inbox, and read emails. This massive performance improvement was achieved with an entire technical rebuild of the mobile app. Now, your local device efficiently caches information about your inbox, enabling you much more efficiently, read and manage your emails.

The new app supports a powerful and fully functional offline mode. For skiff users, who may not always have an Internet connection, have lower Internet speeds, or are simply trying to minimize data usage, offline mode is a critical feature, and is now available to everyone.

You can even keep using the app offline, including reading, emails, replying, labels, or sending messages to different folders. All the updates you make, while offline will later be synced to your web inbox.

We still have a lot more progress to make on the new mobile apps. This includes supporting multiple accounts, a highly requested feature, and even more performance improvements. Please keep in touch with us if you have any questions or feature requests; our team is available on 

Yes. Any account can access and use the new Skiff Mail mobile app

How can you download the new Mail mobile app?

You can download the new Skiff Mail mobile app for iOS and android by visiting 

We will be adding APK downloads to our website soon, you can contact support@skiff.org for an APK file to download.

New Skiff Mail mobile apps launch

 in May 2022, hundreds of thousands of new users have tried our privacy-first, end-to-end encrypted email product. As users start the journey to private email, we've heard countless requests for better features to manage, organize, and sort your incoming emails.

We're excited to announce that mail filters - one of the most powerful ways of staying organized - are released. Filters enable you to automatically keep your inbox, sorted, organized by labels or folders, and tidy from spam or unwanted messages. Whether you've been experiencing annoying, inbound, spam, or frequent repeated messages from a single sender, set up a Mail filter to never be bothered again or to keep track of your most important communications.

To start, navigate to the Skiff Mail app on a desktop web browser, or on the desktop app. You can view your Mail filters from the Skiff Mail, mobile apps, but only create them on the web. Next, open settings from the left sidebar. Then, choose the filters option on the left.

To start, click the “Create new filter” button.

You'll then be able to give your mail filter a name, like “move emails to spam,” "recruiting, emails,” or “sort, custom domain email.”

You can then add filter conditions. Click add condition and choose from a “from” address condition or a “to” address condition, with more coming soon.

Then, choose a set of actions for your mail filter. This includes applying a label, moving mail into a folder, marking it, as read, or unread, or marking it as spam, inbox, or trash.

Mail filters are a powerful feature that are available on every skiff plan, including the free tier. However, the free tier is limited to only two mail filters.

On the Essential, Pro, and Business tiers, you can access 

. Each paid plan also offers numerous additional features, including extra email, aliases, more folders and labels, custom domains, and additional storage space.

Check out our video tutorial on how to create filters 

Are mail filters available on the free plan?

Yes! Every Skiff user can use mail filters for free. However, on the free plan, you will be limited to two mail filters scroll up to see the different plans that are available, including the business plan, which offers how many deadmau5 filtering for any of your aliases, folders, and labels.

Yes. Any Skiff Mail mobile app user can navigate to settings and view mail filters. However, the mobile filter settings are currently limited to view only; you are required to navigate to the desktop or web app to set up new email filters.

What kind of actions can you perform in mail filters?

Mail filters are a powerful tool for keeping your inbox organized. You can easily use mail filters to mark mail as read, apply labels, move, mail to folders, or move them into your spam, trash, or archive inboxes.

Mail filters released

Import from Gmail, now with only one click inside your Skiff inbox.

Skiff Mail is designed to bring total privacy and security to your inbox, with end-to-end encrypted messages, tracker blocking and protection, and aliases for owning your identity. Today, your journey to privacy becomes even simpler with one-click migration from Gmail.

 inside Skiff Mail. Then, click “import” next to the Gmail icon. You’ll be prompted with a modal to “Sign in with Google,” and you will be redirected to a page to sign in with your Google Account. After that, click “allow” to grant access to Skiff, and you will be redirected to your Skiff inbox.

At the bottom of your screen, you’ll see an update in the bottom right corner that your emails from Gmail are importing. It may take minutes or hours to import large inboxes, which can have tens or hundreds of thousands of emails.

If you want to disable import, simply visit the import page again and click “disable” next to the Gmail icon.

Your Gmail inbox will continue to sync with your Skiff account until you disable import from the import settings page. This means that new emails sent to your Gmail account will also be sent to your Skiff inbox; they will remain marked as read inside Skiff.

If you want them to be unread inside Skiff, we would suggest enabling forwarding from Gmail to Skiff instead, which will send over new emails to Skiff. You can see instructions on enabling forwarding inside the 

Not yet, but this is coming soon. You’ll be able to apply labels to all imported mail, or categorize it into a folder. If you want to take advantage of this feature, we would suggest waiting to import your emails and joining our 

Yes. Until you disconnect your inbox from the import page, new emails sent to your Gmail account will continue to populate inside your Skiff account.

Yes! If you are importing thousands of emails, it may take time for all of your emails to be properly indexed and searchable. Then, you will be able to search email subject, content, sender, and to/cc/bcc addresses.

Does imported mail count towards my storage quota?

Yes. All imported mail counts towards your storage quota, just like normal received mail.

What happens if I run out of storage space?

If you run out of storage space during the import process, the import will stop, and you will be unable to send or receive emails. You can upgrade to 100 GB of storage on the Pro plan, and 1 TB of storage on the Business plan from the in-app 