Latest news with #privateAI
Fast Company
6 days ago
- Business
- Fast Company
Proton's new Lumo AI is all about privacy
Proton is getting into generative AI with an assistant called Lumo, which it pitches as a more private alternative to ChatGPT. While Lumo will offer a similar chat-based interface with support for web search and file analysis, Proton says it won't store records of users' conversations or use them to train AI. Lumo is available for free on the web and mobile devices, with an optional $13-per-month or $120-per-year subscription for unlimited chats, extended chat history, and larger file uploads. Andy Yen, Proton's founder and CEO, says Lumo is a way for people to utilize AI assistants without having to worry about how their conversations could be used. 'I think it's critically important, given the amount of sensitive information that we are dumping into AI, that there be a private alternative,' Yen says. Why you might want Proton's private AI It's already possible to maintain some privacy while using major AI tools. ChatGPT, for instance, offers a setting to opt out of training OpenAI's models. It also provides a 'Temporary Chat' feature for conversations that don't appear in your chat history or affect what ChatGPT remembers about you. Google's Gemini also lets users opt out of model training through a Gemini Apps Activity setting. But those settings are not the default, and neglecting them effectively sends your data into a black box. Once your data's been fed to a model for training, it can be extremely difficult to remove, and some providers, including Google, will even show a subset of conversations to human reviewers without disclosing when that happens. ChatGPT also warns that its 'Temporary Chat' mode still stores conversations for up to 30 days for safety reasons. 'Before we created Lumo, I'd use ChatGPT sometimes,' Yen says. 'But then I'd feel really dirty after using it, because who the hell knows what Sam [Altman, OpenAI's CEO] is going to do with all my data?' Proton isn't the first company to offer a more private alternative. The privacy-centric search engine DuckDuckGo launched its own AI tool, last year, with a similar promise not to keep a record of users' conversations or use them for AI training. The difference is that DuckDuckGo has arrangements with major AI providers such as OpenAI and Anthropic, and it sends queries to their servers through a proxy that removes personal information. Those AI providers have promised not to use conversations for training purposes. Proton isn't involving major AI providers at all. Instead, Proton's Lumo AI uses a mix of open-source models that the company runs on its own servers. While conversations aren't end-to-end encrypted, Proton says it doesn't keep logs of users' chats, and conversations are decrypted only on users' devices. 'We're giving people a very clear guarantee of privacy in that your chat history is never going to be saved, logged, or even accessible to us, because it's encrypted in a way that we cannot actually decrypt it,' Yen says. And why you might not Proton's privacy-centric approach has trade-offs, both with its other products and with Lumo. With Proton Mail, the company can't offer server-based email search because it has no way of accessing users' email contents. Instead, Proton builds a local search index on each device where the Mail app is stored. I recently moved away from Proton Mail in large part because the search function was too unreliable. In the case of Proton's Lumo AI, its capabilities are already more limited than other assistants. It can't connect to other apps and services—though Yen says integrations with Proton's email service and document editor are possible—and it can't tailor its responses based on past conversations, akin to ChatGPT's ' Memory ' feature. Its mobile app offers voice input, but not a free-flowing voice conversation mode. Proton also hasn't disclosed which open-source models it's using, and they may not be on par with state-of-the-art models from companies like OpenAI and Anthropic. Yen says he hasn't noticed any issues in his own use of Lumo, but acknowledges that some trade-offs are likely. 'If you want to do things in a privacy-first way, there are going to be sometimes compromises that have to be made,' he says. 'It's a new way of doing AI, new ground that we have to break.' What it means for Proton For Proton as an organization, a foray into generative AI could also make some users uneasy. Proton has always prided itself on running a sustainable business that doesn't rely on venture capital or public shareholders, and it now operates as a nonprofit. Generative AI, meanwhile, is famously a money pit, and like other AI providers, Proton will be offering Lumo access for free. More broadly, AI competes on some level with human creativity and employment, and it uses vast amounts of energy. Proton has already faced a backlash from some of its users after adding AI writing tools to its document editor, then invoked a similar backlash days later with a foray into cryptocurrency wallets. Yen's feeling is that AI represents the future of the web and isn't going away, so Proton should offer a private alternative. While Proton will likely lose money at the outset, that was also the case when it entered the VPN and email businesses. Yen believes the organization can operate AI efficiently and can always adjust what it offers for free if the losses pile up. 'We believe strongly that this is the right thing to do for the world at this moment, and we're going to pursue it even if it ends up costing us money,' Yen says. 'But of course we're not going to compromise Proton overall financially.'
Yahoo
09-07-2025
- Business
- Yahoo
The AI Privacy Crisis Has a Solution: Kynismos AI Launches Revolutionary Platform That Makes Your Conversations Truly Invisible
NEW YORK, NY - July 8, 2025 (NEWMEDIAWIRE) - Kynismos AI is challenging the status quo of modern artificial intelligence. Every day, millions of people pour their most intimate thoughts into AI systems that promise to help them think, create, and solve problems. What they don't realize is that every word becomes permanent evidence in corporate databases, accessible to insurers, employers, law enforcement, and data brokers who treat human curiosity as a commodity. For professionals who work with clients, the stakes are even higher: every query into mainstream AI platforms creates potential liability exposure that could threaten careers and client relationships. Today, that changes. Kynismos AI announced the public launch of the world's first truly private AI platform, offering full access to ChatGPT, Claude, and other leading AI models through a revolutionary architecture that makes user surveillance mathematically impossible - even for Kynismos itself. "We're dismantling the surveillance business model that powers today's AI," said Andrew Sispoidis, CEO and co-founder of Kynismos AI. "Your thoughts should belong to you, not become training data for corporations or evidence in someone else's agenda." The Hidden Cost of "Free" AI While major AI platforms promise privacy through policies, the reality is stark: every prompt is logged, every conversation stored, and every interaction builds detailed psychological profiles that follow users for life. Insurance companies purchase "conversational data" to adjust premiums. Employers monitor AI usage. Courts subpoena chat logs as evidence. For professionals, this creates an impossible choice: avoid AI and fall behind competitively, or use it and accept unknown liability exposure. Lawyers worry about ethics violations. Healthcare professionals face regulatory concerns. Consultants risk exposing client strategies. Even paid AI subscriptions continue this data harvesting, because the real business model isn't your monthly fee - it's your digital identity. True Privacy Through Mathematical Certainty Kynismos AI operates on a fundamentally different principle: zero-knowledge architecture that encrypts and anonymizes user interactions before they ever leave a device. While AI models still process user queries to provide responses, Kynismos ensures they can never link those queries to any specific individual. "We didn't just promise privacy - we made surveillance physically impossible," explained Yuval Drori, CTO and co-founder. "The AI sees your question and provides an answer, but it has no idea who asked it. No tracking, no profiling, no permanent records tied to your identity." The technology works by encrypting prompts locally, stripping all identifying information, routing requests through anonymous proxy networks, and ensuring no logs, metadata, or identity traces remain anywhere in the system. Professional-Grade Power, Complete Anonymity Unlike privacy tools that compromise functionality, Kynismos AI offers the full power of leading AI models through an interface as simple as any mainstream platform. Users can access different AI systems to leverage their unique strengths - Claude's reasoning capabilities, ChatGPT's versatility - while maintaining complete anonymity. Rather than forcing users to maintain separate subscriptions to multiple AI services, Kynismos provides access to leading models through a single subscription. The current platform supports real-time access to ChatGPT, Claude, and other leading models, user-controlled model selection, seamless switching between models while maintaining anonymity, and complex query workflows - all with zero compromise on AI capabilities. The company is developing advanced features for subsequent releases, including intelligent model switching, concurrent agentic functions across multiple models, automated redaction, and confidential information protection. "You can access the best AI capabilities available while ensuring that your identity remains completely invisible to surveillance systems," Sispoidis said. "Future versions will enable AI agents to work concurrently across multiple models, delivering capabilities that no single AI platform can match." Beta Access Opens as Privacy Crisis Accelerates Kynismos AI is dramatically expanding its beta program ahead of the official launch at the end of July. The company is offering "Founder's Edition" access to early adopters who want to experience truly private AI before it becomes publicly available. "Every day people wait is another day their digital thoughts become permanent corporate assets," said Laura Bang, Growth Marketing Lead. As AI integrates deeper into daily life and professional practice, the window for establishing privacy-preserving alternatives is closing rapidly. Major AI companies are racing to lock in users and normalize surveillance-based models before privacy concerns reach mainstream consciousness. A Vision for Human-Centered AI "We're proving that technological advancement and human dignity can coexist," Sispoidis said. "The future of AI doesn't have to be built on tracking and profiling. It can be built on respect for human autonomy and the fundamental right to private thought." Kynismos AI's Founder's Edition is available immediately, with full public launch scheduled for the end of July 2025. The platform operates on a subscription model that aligns the company's incentives with user privacy rather than data extraction. "Our customers pay us directly instead of becoming the product," said COO Tina Santana. "It's a revolutionary concept: a technology company that actually serves the people who use it." About Kynismos AI Kynismos AI is a privacy-first artificial intelligence platform that enables users to access leading AI models like ChatGPT, Claude, and Gemini without compromising their personal data or digital privacy. Through our revolutionary zero-knowledge architecture, user interactions remain completely private and anonymous - we literally cannot access, store, or share user data, even if compelled to do so. As AI becomes embedded in every aspect of daily life, Kynismos provides the essential infrastructure for maintaining human autonomy and digital sovereignty in an increasingly surveilled world. Social LinksX | Instagram | YouTube | LinkedIn Media ContactFull Name: Laura BangTitle: Growth Marketing LeadCompany Name: Kynismos AIEmail: media@ Error in retrieving data Sign in to access your portfolio Error in retrieving data Error in retrieving data Error in retrieving data Error in retrieving data
