Xayn Starts Beta Program of Personalised and Private News Assistant
New Beta app offers continuous clutter-free news stream
- Xayn blends personalised news recommendations with privacy based on edge AI
- The Xayn News Assistant is powered by the Xayn search engine
- Individualised personalisation protects from polarisation by recommending news stories only based on unique user preferences and not on popularity in peer group
Berlin, 09. March 2022 – Xayn, the privacy tech company, starts today the beta program of their news stream that delivers users personalised news and blog content in a continuous full screen format. This is the latest development of the company’s original search engine launched in December 2020. With its self-learning feed, the new app targets digital natives who care about great journalism more than any one media brand and who define privacy as a human right. The app blocks cookies and ads to create a private seamless reading experience. This boutique assistant for news offers users diverse, easy, and secure access to relevant information, all while protecting their privacy and autonomy.
The beta app is powered by Xayn’s search engine based on edge AI that protects users’ privacy by leaving all personal data on their smartphones. This individualised personalisation based on decentralised technology also protects users from polarisation caused by filter bubbles. All recommendations are based on the unique user behaviour instead of any popularity-based algorithms or cross-profiling which are known to fuel polarisation on social media platforms. On social media, content that gets a lot of engagement from many people is ranked higher in the individual personal profiles. In addition, the AI assistant constantly explores new user behaviours and tries out new things to avoid a too narrow one-sidedness, since a user’s interests might change over time.
“Today’s internet is filled with an abundance of information which makes it hard to find the useful one. To add insult to injury, users are used for their data and lose their privacy,” says Leif-Nissen Lundbæk, Co-Founder and CEO of Xayn. “We want to change this and, therefore, develop this news assistant to give users the most seamless and privacy-protected news reading experience. Now, we want to get our users on board with our new beta program so that we can develop the app according to their wishes and needs. We are convinced that privacy-friendly technology will only succeed if it takes users and their needs seriously.”
Highlights and roadmap of the Xayn News Assistant
Highlights of the app in beta include among others:
- Curation that cuts through the noise: Multiple AI models prevent users from spiralling down ever-narrower echo chambers while responding in real-time to their preferences
- Clutter-free reading experience: Xayn blocks all ads and pop-ups for an easier reading experience
- User-controlled like and dislike features: A one-click like or dislike function keeps users in direct control of training the algorithm
- No data honeypot: Personal user data will never be compromised because it is never transferred to a honeypot server – all personal data stays on the phone
Initially, the beta version of our Xayn News Assistant will be released as a basic version, free of charge. During the beta program, Xayn will continuously introduce new features which users can test extensively and on which they can give feedback. Beta users will be the first to know about new product offerings and will maintain their unlimited free access to the app once it’s been released in the official freemium version later this year. Once the official Xayn News Assistant is released, users can try it out for free for seven days. Thereafter, they can continue to enjoy individualised personalisation and a full protection of their digital privacy in the Xayn News Assistant for only 2.99 EURO a month.
We firmly believe that a free and independent press is an important pillar for strong democratic societies. Our technology supports the human right of privacy on the web. With our Xayn News Assistant, we extend this support to the right to a free press, so that good journalism can continue to flourish in the digital age. This is why after the successful beta program and the official product release later this year, Xayn aims to partner with the media industry so that they can reach younger target audiences with their high-quality journalistic content. Starting today, users can download the Xayn News Assistant in beta for Android, the iOS version will follow in the next weeks. For more information on the beta program and to download the app for Android and iOS go to: beta.xayn.com
The Berlin-based company Xayn provides users with a privacy-protected gateway to the Internet. Their private search engine presents users with personalised search results and a curated news stream – without ads or data collection. Xayn is based on decentralised edge AI made in Europe and ushers in a new generation of user-friendly privacy tech.
The AI company started as a research project at The University of Oxford and Imperial College London by Leif-Nissen Lundbæk (PhD) and Professor Michael Huth. Together with Felix Hahmann, they founded the tech company in 2017. To this day, that academic vision remains with a workforce comprised of around 30% PhDs. The company also developed XayNet, the open-source framework for Masked Federated Learning. The start-up has received investment funding of 19.5 million EURO from Global Brain Corporation, KDDI Open Innovation Fund, Earlybird as well as Dominik Schiener. Xayn has worked with corporations such as Porsche, Daimler, Deutsche Bahn, and Siemens.
Downloads & Links
- Xayn Search for iOS and Android
- Xayn News Assistant Beta for Android
- Pictures (product, management, office)
- Company factsheet, boilerplate, management bios
1. How does the technology for the Xayn News Assistant work?
For us, a privacy app is one that does not send any of the personal data (IP address, search requests, and so forth) from the app elsewhere for storage or processing. “What happens within the app, stays within the app.” We achieve this with decentralized or device-level computing, making it impossible for us to access personal data of our users.
Our News Assistant is a search assistant; it finds interesting articles from the internet and curates them for users in a personalised stream while protecting their privacy. To tailor the content stream to user preferences, we built an AI that learns from user interactions (likes, dislikes and time users spend with each document) directly on their device. Thus, all their personal data always stays only on their device and is never sent to us.
On each device, different AI models work together, creating an AI assistant that performs multiple jobs to search the Internet for content most relevant to a specific user:
- The assistant first analyses the semantics of news and articles and represents them mathematically, so that the assistant can measure the semantic similarities of texts.
- It then measures the individual user interactions with the respective articles directly on the device side, keeping this user data private.
- In case a user likes a document, the assistant also generates key phrases from the document.
- Further, the assistant combines the semantics of a document with the user interactions to generate a mathematical personal model of a user; this model is kept only on the device side. The more a user interacts with the app, the easier it is for the AI models to figure out what the preferences of the user are. To generate the content stream, the assistant uses the extracted key phrases of liked documents to specifically search for more documents in an index that consists of billions of articles; this index gets updated all the time with newly added content.
- The assistant performs such searches perpetually after a couple of articles have been viewed by the user, generating a ceaseless stack of content displayed in the app.
- The assistant uses the personal model that it consistently updates to determine the order in which documents from this stack are shown to the user.
- Lastly, the AI assistant constantly explores new user behaviours and tries out new things to avoid a too narrow one-sidedness, since a user’s interests might change over time.
We forward the user’s search requests which are generated automatically by the Xayn AI for the news feed between their app and our Xayn content index, and vice versa. The index only includes URL and title of the articles which we use for the reranking. Whenever a user clicks on a document, the reader mode is activated and processes locally.
We must do this to provide users with the respective content. This index does neither log, store nor even analyse any of the user´s personal data and acts as a stateless service. The user’s personal data is not used at all once forwarded between their app and the index, and vice versa. In addition, we encrypt the information that travels from the app to the index and back and only those endpoints can decrypt this.
Because we believe that a transparent Internet is a better Internet and wanted to give back to the developer community, we also open-sourced our entire codebase – including our AI-based assistant, the app itself, our entire infrastructure as well as all additional components, such as the reader mode. You can find all of our repos at Github.
2. How do we protect our users from polarisation caused by filter bubbles?
Our individualised personalisation is based on a decentralised technology used in our Xayn News Assistant. This technology also protects users from polarisation caused by filter bubbles. This is so since all recommendations are based on the unique user behaviour instead of any popularity-based algorithms or any cross-profiling of users, which are known to fuel polarisation on social media platforms. On social media, content that gets a lot of engagement from many people is ranked higher in the individual personal profiles. This means that the content suggestions are based largely on the community behaviour and not so much on the unique behaviour of each individual user.
The Xayn News Assistant, in contrast, generates article suggestions and their rankings based solely on the behaviour of the user who consumes and interacts with this private, personalized news stream. Our app also doesn’t have any community element to it. Such elements make up large parts of social media. The Xayn News Assistant offers no commenting, no sharing of articles or images with other users, no interaction with other users. This means users can enjoy news in a completely private way.
Our business model is not based upon ads and/or collecting user data – we don’t have access to any personal user data. Therefore, our users can be sure that when using our app, they won’t be micro-targeted by any other company, institution, or organisation based on user interactions with the Xayn News Assistant.
The Xayn AI contains an exploratory part that delivers users new and perhaps surprising content suggestions. This achieves two aims: to allow for shifts of user interests in content over time, and to occasionally expose users to content they might not ordinarily expect to see. This is realized with what is called a Contextual Multi-Armed Bandit as one of our AI models.
3. Can users read paywalled content in the Xayn News Assistant?
The Xayn News Assistant trains directly on device-level to create a personalised news stream for each individual user. This also means that users might see recommendations in their curated news feed that link to content behind paywalls. In this first stage of our Xayn News Assistant, users would need the respective subscriptions to access these paywalled articles. Media publishers benefit from this as more potential users might be led to their news content, which will thus receive more attention and respective clicks.
Our app only shows content that isn't behind paywalls and curates it directly on device-level to create a personalised news stream. We believe that the Xayn News Assistant can also be of great interest for the media industry to reach young digital natives who are more and more difficult to reach via traditional media platforms as several studies have shown (see also 3.1.).
We firmly believe that a free and independent press is an important pillar for strong democratic societies. Our technology supports the human right of privacy on the web. With our Xayn News Assistant, we extend this support to the right to a free press, so that good journalism can continue to flourish in the digital age.
Therefore, we’re looking forward to partnering with the media industry – for example, so that users will also be able to access premium content through our Xayn News Assistant.
If you’re a publisher and interested in partnering with us, please contact us via firstname.lastname@example.org.
4. Are we open to partnering with the media industry (e.g., with publishers)?
Yes, we are looking forward to partnering with the media industry. We firmly believe that a free and independent press is without a doubt an important pillar for a strong democratic society – and we want to put our technology which supports human right of online privacy, to the best use.
With our Xayn News Assistant, we want to help media companies reach the younger audience that favours social media over traditional news platforms for access news, as several studies show.
The Digital News Report 2021 by the Reuters Institute for the Study of Journalism at the University of Oxford has shown that those “aged 18–24 (so-called Generation Z) have an even weaker connection with websites and apps (18%) and are almost twice as likely to prefer to access news via social media (34%), aggregators, or mobile alerts.”
The US polling service Morning Consult/The Hollywood Reporter analysed in 2021 that social media is by far the most popular platform for daily news consumption among those aged 18-34 (44%) and those aged 35-44 (40%). Online-only news site land in second in the youngest age group (16%) followed by cable news (15%), radio (15%), and network news (12%).
The new report How Young People Consume News and the Implications for Mainstream Media by the Reuters Institute for the Study of Journalism at the University of Oxford has concluded that
- “young people are looking for content and platforms that allow them to seamlessly flow through the digital space in an uninterrupted way”
- “Social Media has an almost unrivalled relevance as it allows a less mediated experience. It also offers opportunities to control and curate your own unique news experiences”
- “the experience of news should feel as easy and accessible as Facebook and Netflix. This is partly about how the content is written and presented (…), but also about how relevant and interesting content is surfaced without having to work for it. Instant, frictionless access, recommendations that feel relevant and useful, and the right tone of voice will be critical to building loyalty and trust with these groups”
With our Xayn News Assistant, we cater for these preferences of the younger audience and combine them with an AI-based personalised and private curation of quality journalism. The look and feel of our app are like popular social media platforms such as TikTok or Instagram, making it more attractive to digital natives. Our decentralised edge AI also always protects the privacy of the user.
We believe that our app can be of great interest for the media industry to reach young digital natives. Therefore, we’re looking forward to partnering with the media industry. With our Xayn News Assistant, we extend our support for privacy as a human right to the right to a free press, so that good journalism can continue to flourish in the digital age.
If you’re a publisher and interested in partnering with us, please contact us via email@example.com.
5. What about copyright?
The Xayn News Assistant includes browser components and creates a continuous, personalised and private feed for news, articles, and blog content. That feed only shows headlines and pictures but no snippets. Every time a user clicks on a news suggestion that appears in their feed, the owner of the website receives the web traffic. We block the automated user profiling by blocking cookies and third-party tracking in the reader mode.
Users can only read content that also can be crawled via robot.txt. This means that they cannot read content that is behind paywalls when using the Xayn News Assistant. Each publisher can protect all their content behind technical walls. When using Xayn, the privacy of the user is protected because we block cookies, ads and third-party tracking in the reader mode. Reader modes are standard in privacy settings of many browsers (see Mozilla’s definition of a browser) such as Safari, Chrome, Firefox, and Brave. Any user who wishes to browse privately, can change these settings. With the Xayn News Assistant, each user can read news content in a reader mode and – with a simple click – can chose to open an article in another browser of their choice.
Our Xayn News assistant therefore only sorts through content from the Internet to give users personalised recommendations, leading them to different websites. This way, Xayn enables users to access a website found by Xayn’s decentralized and personalized AI. Therefore, the Xayn News Assistant and its usage support compliance with the copyrights of content providers.
We believe that our model can also be of great interest for the media industry to reach the younger audience who favours social media over traditional news platforms for access news, as several studies show (see also 3.1.). Therefore, we’re looking forward to partnering with the media industry – for example so that users will also be able to access premium content. We want to create new ways of financing the media industry so that good journalism can still flourish by getting more attention and reaching a broader audience. With our Xayn News Assistant, we extend our support for privacy as a human right to the right to a free press, so that good journalism can continue to flourish in the digital age.
If you’re a publisher and interested in partnering with us, contact us via firstname.lastname@example.org.
Dr. Clara Herdeanu
Head of Communications