Proton's Andy Yen Says It Will Have To Proton's CEO is a champion of encryption for everyone. So why is he going all in on un-encryptable AI? In the eyes of many privacy advocates, AI is just another surveillance technology . It supercharges online tracking and real-world spying tools, sifts through vast troves of everyone's data to spit out its results, and then encourages users to share their deepest secrets and desires with a cloud-hosted large-language model. Surveillance capitalism made AI possible, and AI returns the favor by making surveillance more powerful and the businesses that run on it more data-rich and profitable than ever. Andy Yen is among the most influential voices in the tech industry who equates AI with privacy invasion. He says the push to integrate AI into every consumer tech product--whether people want it or not--is driving new users in droves to Proton's products, which are, put simply, the end-to-end encrypted versions of every Google service from Gmail to Google Docs, Sheets, Calendar, and Meet. Yet Yen is anti-AI, he says, any more than he is anti-internet. In late June, Proton released the latest version of its very own AI chatbot, Lumo--just one way that Proton is integrating AI ever more deeply into its suite of tools, but seeking to do so in a way that's more voluntary and preserves users' privacy far more than the typical tech giant. That willingness to embrace the apparent contradiction between AI and privacy is just one way in which Yen's views don't cut cleanly across party lines--from current US politics to the lessons of Edward Snowden's leaks to the question of the best corporate structure for a tech company that's not solely designed to maximize profit. You can read our conversation about all of it here or listen wherever you get your podcasts. ANDY GREENBERG: I want to start by asking about your background and the origin story of Proton. You're a trained particle physicist. You worked at CERN, the nuclear research facility, but you were inspired by the leaks of Edward Snowden back in 2013 to create Proton, right? I'm not actually a tech person or even an entrepreneur by training, so I really happened into this field a little bit by accident. I used to work at the Large Hadron Collider, which is a particle accelerator.

When you purchase through links in our articles, we may earn a small commission. Gmail offers fully encrypted email. Your email isn't as private as you might assume--and you should look outside Google if you want to change that. Email that only your recipient can read--sounds like a great idea for sensitive messages, right? Especially if you could have that within Gmail.

Lumo 2.0 can also search for relevant background information. Proton has rolled out its biggest update yet for the Lumo chatbot, almost a year after it launched . Lumo version 2.0 now comes with image recognition and generation, finally making it a legitimate competitor to ChatGPT and Gemini. Proton says the updated chatbot has the capability to generate images, as well as to analyze and edit them. Conversations involving images are still protected by zero-access encryption like all chats on Lumo, which means they can only be accessed on your device.

Proton's Lumo AI chatbot now has an encrypted space for your projects Lumo 1.3 is now available to all users. Proton launches Projects within Lumo. Proton's latest update for Lumo, its privacy-focused chatbot, introduces a feature called Projects. It's a dedicated and encrypted space for tasks that you know you'll access again and again over an extended period of time, such as papers you'll have to work on the whole semester or plans for a big trip you're taking later this year. Lumo will remember and keep all the information and all the files you upload for every project you create. Any document you upload or resources you add to the chat will sync across devices, so you don't have to repeat yourself every time you access a task.

Ear biometrics offer a stable and contactless modality for identity recognition, yet their effectiveness remains limited by the scarcity of annotated data and significant intra-class variability. Existing methods typically extract identity features from individual impressions in isolation, restricting their ability to capture consistent and discriminative representations. To overcome these limitations, a few-shot learning framework, ProtoN, is proposed to jointly process multiple impressions of an identity using a graph-based approach. Each impression is represented as a node in a class-specific graph, alongside a learnable prototype node that encodes identity-level information. This graph is processed by a Prototype Graph Neural Network (PGNN) layer, specifically designed to refine both impression and prototype representations through a dual-path message-passing mechanism. To further enhance discriminative power, the PGNN incorporates a cross-graph prototype alignment strategy that improves class separability by enforcing intra-class compactness while maintaining inter-class distinction. Additionally, a hybrid loss function is employed to balance episodic and global classification objectives, thereby improving the overall structure of the embedding space. Extensive experiments on five benchmark ear datasets demonstrate that ProtoN achieves state-of-the-art performance, with Rank-1 identification accuracy of up to 99.60% and an Equal Error Rate (EER) as low as 0.025, showing the effectiveness for few-shot ear recognition under limited data conditions.

Developing robust representations of chemical structures that enable models to learn topological inductive biases is challenging. In this manuscript, we present a representation of atomistic systems. We begin by proving that our representation satisfies all structural, geometric, efficiency, and generalizability constraints. Afterward, we provide a general algorithm to encode any atomistic system. Finally, we report performance comparable to state-of-the-art methods on numerous tasks.

Researchers have developed a method for diagnosing autism which could spare families years of uncertainty and spur crucial earlier treatments. The new AI analysis can identify the genetic markers of autism via biological activity in the brain, they report, with 89 to 95 percent accuracy. This new method starts out with standard brain-mapping via magnetic resonance imaging (MRI) before re-analyzing those scans via AI to detect the movements of proteins, nutrients and other processes within the brain that may indicate autism. 'Autism is traditionally diagnosed behaviorally,' via a person's speech, for example, as the medical team behind the process noted. 'But [it] has a strong genetic basis.'

Proton Mail has a new AI-powered feature that could help it keep pace with the artificial intelligence tools Google and Microsoft offer for their email services. Proton Scribe is an AI writing assistant that can help you compose and clean up your drafts. Scribe was designed with privacy in mind -- the assistant can't train on your inbox data, as Proton Mail has a zero-access approach to encryption. Proton doesn't save or log anything from your email drafts either. According to Proton, a writing assistant was one of the most-requested features in a recent user survey.

The Alpha Magnetic Spectrometer (AMS) is a high-precision particle detector onboard the International Space Station containing six different subdetectors. The Transition Radiation Detector and Electromagnetic Calorimeter (ECAL) are used to separate electrons/positrons from the abundant cosmic-ray proton background. The positron flux measured in space by AMS falls with a power law which unexpectedly softens above 25 GeV and then hardens above 280 GeV. Several theoretical models try to explain these phenomena, and a purer measurement of positrons at higher energies is needed to help test them. The currently used methods to reject the proton background at high energies involve extrapolating shower features from the ECAL to use as inputs for boosted decision tree and likelihood classifiers. We present a new approach for particle identification with the AMS ECAL using deep learning (DL). By taking the energy deposition within all the ECAL cells as an input and treating them as pixels in an image-like format, we train an MLP, a CNN, and multiple ResNets and Convolutional vision Transformers (CvTs) as shower classifiers. Proton rejection performance is evaluated using Monte Carlo (MC) events and ISS data separately. For MC, using events with a reconstructed energy between 0.2 - 2 TeV, at 90% electron accuracy, the proton rejection power of our CvT model is more than 5 times that of the other DL models. Similarly, for ISS data with a reconstructed energy between 50 - 70 GeV, the proton rejection power of our CvT model is more than 2.5 times that of the other DL models.

Image decomposition plays a crucial role in various computer vision tasks, enabling the analysis and manipulation of visual content at a fundamental level. Overlapping images, which occur when multiple objects or scenes partially occlude each other, pose unique challenges for decomposition algorithms. The task intensifies when working with sparse images, where the scarcity of meaningful information complicates the precise extraction of components. This paper presents a solution that leverages the power of deep learning to accurately extract individual objects within multi-dimensional overlapping-sparse images, with a direct application in high-energy physics with decomposition of overlaid elementary particles obtained from imaging detectors. In particular, the proposed approach tackles a highly complex yet unsolved problem: identifying and measuring independent particles at the vertex of neutrino interactions, where one expects to observe detector images with multiple indiscernible overlapping charged particles. By decomposing the image of the detector activity at the vertex through deep learning, it is possible to infer the kinematic parameters of the identified low-momentum particles - which otherwise would remain neglected - and enhance the reconstructed energy resolution of the neutrino event. We also present an additional step - that can be tuned directly on detector data - combining the above method with a fully-differentiable generative model to improve the image decomposition further and, consequently, the resolution of the measured parameters, achieving unprecedented results. This improvement is crucial for precisely measuring the parameters that govern neutrino flavour oscillations and searching for asymmetries between matter and antimatter.