• February 2, 2025: Tinfoil Team launched Tinfoil Private Chat with DeepSeek R1-70B, aiming to redefine AI trust by replacing blind faith with provable security.

  • Current situation: Since R1 was released last week, there's been an increase in concerns about AI and privacy. Even OpenAI employees are worried about user data going to AI and cloud providers.
  • Problem: Models like DeepSeek R1 are too large to run locally, so users need to trust multiple third parties that can access, collect, sell, and use their private data.
• Replace trust with provable security: Tinfoil is using confidential computing with additional layers of hardware security, code transparency, and auditability to ensure only end-users access private data. It guarantees that no one else can access the data, providing the highest levels of security and privacy in cloud computing.
• What else is happening at Tinfoil: The private chat and inference endpoint are just the beginning. Tinfoil is building an entire platform to deploy private, verifiable versions of AI applications like content moderation tools, secure code editors, and more.
• Subscribe for Updates: Stay updated with Tinfoil's latest blog posts and announcements.

Footnotes:

1. Recent large-scale attacks on US telecommunications systems led to the recommendation of using end-to-end encrypted messaging apps.
2. A standard enterprise contract with OpenAI only prohibits training on user data, not accessing or analyzing it.
3. Complete end-to-end transparency and client-side verification is crucial to alleviate concerns about backdoors, as highlighted by security researchers. More blog posts are coming soon.