Moxie Marlinspike, the co-founder of Signal, has launched a new AI project aimed at addressing growing concerns about privacy in conversational AI. The service, called Confer, is designed as a privacy-focused AI assistant and an alternative to tools like ChatGPT and Claude, offering similar chat-based interactions without data collection, model training on user conversations, or ad targeting.

The project, which launched in December, arrives as AI assistants become more deeply embedded in everyday workflows—and as companies experiment with advertising and data-driven monetization models. For users wary of sharing sensitive personal information with AI systems, Confer offers a fundamentally different approach.

Marlinspike says the project is a direct response to the uniquely personal nature of conversational AI. Unlike traditional software, chat-based systems actively invite users to share private thoughts, personal concerns, and sensitive information.

“It’s a form of technology that actively invites confession,” Marlinspike said. “Chat interfaces like ChatGPT know more about people than any other technology before. When you combine that with advertising, it’s like someone paying your therapist to convince you to buy something.”

Confer is designed so that this kind of data access is structurally impossible.

Rather than relying on trust policies or data-handling promises, Confer uses a layered technical architecture intended to prevent access to user conversations entirely.

Messages are encrypted using the WebAuthn passkey system, securing communications between the user and the service. On the server side, AI inference runs inside a Trusted Execution Environment (TEE), combined with remote attestation to verify that the system has not been altered or compromised.

Within that environment, Confer routes queries to a set of open-weight foundation models, allowing the system to generate responses without storing or exposing conversation data.

While the setup is more complex than a typical AI inference pipeline, Marlinspike argues it is necessary to deliver on the service’s core privacy guarantees.

Confer’s emphasis on privacy comes with constraints. The free tier is limited to 20 messages per day and five active chats. A paid plan, priced at $35 per month, removes usage limits and offers unlimited access to more advanced models and personalization features.

That price is significantly higher than ChatGPT Plus, which currently costs less—but Marlinspike is explicit about the tradeoff. In this model, privacy is treated as a core system requirement rather than a feature layered on top of a data-driven service.

Q: Who built Confer?
A: Confer was created by Moxie Marlinspike, co-founder of the encrypted messaging app Signal.

Q: Can Confer access user conversations?
A: No. Confer is architected so that conversations cannot be accessed by the service operator or used for training or advertising.

Q: How is Confer different from ChatGPT or Claude?
A: While the interface is similar, Confer uses encryption, trusted execution environments, and open-weight models to prevent data collection.

Q: How much does Confer cost?
A: The free tier is limited. Paid access costs $35 per month.

Confer highlights a growing divide in the AI ecosystem. As mainstream AI assistants explore advertising and data-driven business models, privacy-first alternatives are emerging that treat confidentiality as a foundational architectural constraint rather than a policy choice.

The tradeoff is clear. Services like Confer are more technically complex and more expensive, but they appeal to users who see conversational AI as deeply personal—and who are unwilling to exchange that intimacy for convenience or lower cost. For users who treat AI assistants as tools for reflection, planning, or sensitive work, privacy-by-design architecture may outweigh ease of use or price.

At the same time, Confer’s approach underscores why privacy-by-design has been difficult to implement at large scale. Systems that fully encrypt conversations and limit server-side visibility reduce opportunities for data collection—but they also constrain safety monitoring, abuse detection, and rapid iteration. In this sense, Confer and platforms like ChatGPT reflect different governance philosophies: Marlinspike prioritizes accepting limits to avoid surveillance, while OpenAI prioritizes visibility to manage risk and accountability at global scale.

As AI assistants continue evolving into long-term companions for work and personal life, Confer raises a broader question: whether privacy in AI will remain a premium option for a subset of users, or become an expectation that more people increasingly demand.

Editor’s Note: This article was created by Alicia Shapiro, CMO of AiNews.com, with writing, image, and idea-generation support from ChatGPT, an AI assistant. However, the final perspective and editorial choices are solely Alicia Shapiro’s. Special thanks to ChatGPT for assistance with research and editorial support in crafting this article.