Signal co-founder Moxie Marlinspike launched a new project in December called Confer, a privacy-conscious alternative to AI personal assistants like ChatGPT and Claude. Confer aims to provide a similar user experience to these popular chatbots, but with a backend designed to avoid data collection, addressing growing privacy concerns surrounding AI services.

The impetus behind Confer stems from the intimate nature of AI chatbots and the potential for misuse of personal data. Marlinspike argues that these interfaces "actively invite confession" and can collect more personal information than any previous technology. He expressed concern about the potential for advertising companies to exploit this data, comparing it to "someone paying your therapist to convince you to buy something."

Unlike many AI models, Confer's architecture ensures that user conversations cannot be used to train the model or target advertisements. This is achieved by preventing the host from ever gaining access to the conversation data. The open-source nature of the project, similar to Signal, is intended to foster trust and transparency.

The rise of AI personal assistants has sparked debate about data privacy. Many users are wary of sharing personal information with AI models, fearing that it will be retained and potentially used for advertising or other purposes. OpenAI, the company behind ChatGPT, is already testing advertising, raising concerns that the data collection practices of companies like Facebook and Google could become commonplace in chatbot interactions.

Confer represents an attempt to address these concerns by offering a privacy-focused alternative. By prioritizing data protection and transparency, the project hopes to demonstrate that it is possible to develop AI services that respect user privacy. The project is still in its early stages, and its long-term success will depend on its ability to attract users and developers who share its commitment to privacy.