June 29 – July 5 (Published July 8th)
PERSPECTIVES by Steve Payne
14 Crypto Private Financings Raised: $100M
Rolling 3-Month Average: $400M
Rolling 52-Week Average: $347M
Deals Over $50M: 1
The only notable deal in last week’s holiday-shortened list was Venice.ai’s $65M Series A, led by Dragonfly, at a $1B post-money valuation.
Venice started two years ago to solve the privacy issue with AI, and boasts 3.5M registered users and a $70M annualized run rate. The press release for the financing states the problem:
“More of how people reason, create, and decide now runs through AI. That makes the AI layer the most sensitive surface in a person’s digital life, and most providers treat it as data to keep. Major AI providers store user data permanently. Every prompt is logged, analyzed, and tied to the user’s identity. That record can be sold, hacked, subpoenaed, or handed to a government. As AI becomes the primary gateway to the digital world, that is not a footnote on privacy. It is surveillance aimed at the most personal thing a person has: their thoughts. ‘Intelligence, the lifeblood of civilizational advancement, is becoming a collaboration between man and machine,’ said Erik Voorhees, founder and CEO of Venice. ‘Venice’s mission is to protect it from mass surveillance and censorship.’”
Venice’s solution is architectural rather than policy-based: it markets itself as an “AI safety company,” framing surveillance of users’ data – rather than the content of their prompts – as the greater danger. Conversations are stored on the user’s device rather than on Venice’s servers, and Venice says it does not log prompts. For queries routed to third-party models from OpenAI, Anthropic, xAI, and Google, a proxy obscures the user’s IP address, account, and session data. Users stake Venice’s VVV tokens to earn perpetual daily compute rights (measured in credits called DIEM).
Venice’s approach is not without detractors (security experts among them) or competitors. There are three main approaches to the privacy problem:
- 1. Encrypted/zero-knowledge cloud AI – closest to Venice’s model. Includes Proton’s Lumo, Maple AI, and Kagi.
- 2. Pure on-device/local inference (architecturally stronger privacy, since data never leaves the device). Includes PocketLLM, Private LLM, LLM Farm, MLC Chat, Ollama, LM Studio, Jan, GPT4All, Atomic Chat, Open WebUI, InnerZero, and more.
- 3. Enterprise-grade private deployment – e.g., Claude Enterprise and similar offerings, which compete on contractual/policy-based privacy (no training on inputs, data retention controls, large context windows) rather than architecture.
