Redpill delivers AI privacy by design. All workloads execute in secure hardware enclaves — every LLM query comes with a cryptographic proof so you never have to trust us blindly. Integrate easily via our simple SDK / API.
I'm Marvin, founder of Redpill. We're a small, privacy-obsessed team building the VPN for AI privacy — and I’m thrilled to finally share what we’ve been working on.
💡 Why did we build this?
We built Redpill out of frustration with how every AI interaction today leaks data — to the model provider, the infrastructure host, or both. AI tools have become essential, but the moment you use them, you lose control of your data.
As builders, we knew that privacy shouldn’t rely on “trust us.” So we set out to build a platform where AI privacy is verifiable, not promised.
🧠 What makes Redpill unique?
Redpill is the 1st ever private-first AI platform that runs all workloads in secure hardware (TEE) — meaning every AI query comes with cryptographic proof that your data stays private.
We support 200+ popular models, all protected by a TEE proxy. So you can use the tools you love — GPT, Claude, Gemini, and more — inside a verifiably private environment, through our encrypted chat, SDK, and API.
How does RedPill handle prompts or data that must leave the enclave (e.g., for billing, analytics, or tool use), and what controls exist to prevent accidental exfiltration?
Hi, very cool. Will you provide the latest Qwen3 VL models on TEE GPUs any time soon?
Sorry if I’m not completely understanding, but you need to have subscriptions to be AI services separately or do they come with the software?
@marvin_tong how do you use GPT-5 or Claude when they are not running in a TEE? You cannot pass the data encrypted to them no?
Can definitely see this being used in regulated industries like healthcare and finance. Congrats with the launch!
This looks awesome. How easy is it to plug Redpill into existing tools and workflows?
@marvin_tong @mout The TEE (Trusted Execution Environment) adds roughly 20–40 ms extra per request for attestation and encrypted streaming, depending on GPU type and model size.
@marvin_tong Lots of our clients ask about privacy, it's promising for me! How about the latency? Will it slow down the response time?
That’s a strong stance on AI trust. Does it support on-prem deployments too?
I have been thinking of something similar to the concept of 'VPN for AI,' and now it comes!
As a daily VPN user, I know how it works. But as to VPN for AI, I wonder what latency and cost overheads I should expect compared with calling a model provider directly?
And is there a public SDK/API spec or demo repo I can try, and are there any open-source components I can audit before onboarding?
Hey Product Hunt! 👋
I'm Marvin, founder of Redpill. We're a small, privacy-obsessed team building the VPN for AI privacy — and I’m thrilled to finally share what we’ve been working on.
💡 Why did we build this?
We built Redpill out of frustration with how every AI interaction today leaks data — to the model provider, the infrastructure host, or both. AI tools have become essential, but the moment you use them, you lose control of your data.
As builders, we knew that privacy shouldn’t rely on “trust us.” So we set out to build a platform where AI privacy is verifiable, not promised.
🧠 What makes Redpill unique?
Redpill is the 1st ever private-first AI platform that runs all workloads in secure hardware (TEE) — meaning every AI query comes with cryptographic proof that your data stays private.
We support 200+ popular models, all protected by a TEE proxy. So you can use the tools you love — GPT, Claude, Gemini, and more — inside a verifiably private environment, through our encrypted chat, SDK, and API.