Moxie Marlinspike launched Confer, an AI chatbot designed to protect user privacy, after growing concerns about data collection by mainstream AI services. The tool uses cryptography to prevent the service’s own servers from logging or surveilling user conversations.
"Confer is designed to be a service where you can explore ideas without your own thoughts potentially conspiring against you someday," he wrote in a blog post introducing it.
Mainstream AI chatbots like ChatGPT, Claude, and Gemini are set by default to collect and store private user data, often without restrictions on how it is shared or used.
Matt Green, a privacy- and security-focused computer science professor at Johns Hopkins University, said, "You're giving it this huge profile on you." These models are often used for personal conversations, including topics like health, finances, and relationships.
Marlinspike’s private AI tool is just one of a new generation of services aiming to address privacy concerns with AI. Some services promise not to record conversations, while others anonymize data or use technical guardrails to restrict access to user secrets. The result is a growing number of tools offering varying levels of privacy assurances.
Zero data retention (ZDR) policies are available for enterprise users, but not for average users. OpenAI, Anthropic, and Google offer ZDR for their enterprise versions, which require immediate deletion of user interactions. However, these policies have exceptions, such as Anthropic’s refusal to offer ZDR for its most advanced models due to potential misuse.
Privacy-focused services like Proton’s Lumo and DuckDuckGo’s Duck.ai offer similar promises, but they often relay user requests to other AI models, which can still collect data. Green noted that even stripped of identifying metadata, chat messages can often contain identifiable information.
The growing competition to create less surveillance-prone AI is creating a range of tools, often with confusing assurances for users. As AI becomes more integrated into personal lives, the challenge remains to balance its benefits with the need for privacy protection.
Source: wired
