Gubner

The open-source model with 100% privacy guarantee, secured by Phala’s TEE infrastructure 🛡️ Venice.ai leverages Phala’s TEE-powered GPU infrastructure to run AI inference inside secure enclaves, ensuring that prompts and outputs remain fully private and inaccessible, even to the infrastructure providers. This is confidential AI in action: verifiable, encrypted, and trustless. Don’t expose your data. Explore it yourself: venice.ai phala.com

Venice just went Phala-grade on privacy. Their new TEE mode runs inside hardware-isolated enclaves powered by Phala, with cryptographic proof via remote attestation. Huge congrats to @AskVenice! 🍾🥂

went to meet my tax officer it was just ... everything everywhere all lobster 🦞

Confidential Computing spotted

“Who’s using GPU TEE?” 👇

My Clawdi agent is live 🎥 Now I can use it to automate tasks, generate content, and interact across apps, all powered by the new OpenClaw. With TEE-backed security from @PhalaNetwork, my data and AI workflows run inside secure enclaves, ensuring privacy, integrity, and verifiable execution. Try it yourself: Clawdi.ai🦞

TELUS and Fortanix just launched a Confidential AI solution at @NVIDIAGTC 2026, with cryptographic proof of protection for sensitive AI workloads on sovereign infrastructure. The market is speaking: verifiable, hardware-backed AI security is becoming the standard.

March is already halfway through, and Phala continues to push forward in building the future of confidential AI. 🔒🖥️ The first half of the month brought several important milestones across the ecosystem, from new tools for AI agents to deeper infrastructure integrations and research on real-world AI usage. Here are some highlights 👇 🔴Clawdi Launch Phala introduced Clawdi, bringing the power of OpenClaw agents into a confidential computing environment. By running agent workloads inside Trusted Execution Environments (TEEs), Clawdi enables AI agents to operate with hardware-level privacy and verifiable execution — ensuring that data, prompts, and agent logic remain protected during runtime. 🔵Phala x Intel @PhalaNetwork also announced a collaboration with Intel through Intel Trust Authority, strengthening the attestation layer behind its confidential AI infrastructure. This integration introduces independent verification of TEE environments, reinforcing trust for developers deploying sensitive AI workloads. Confidential AI requires verifiable infrastructure — and this partnership pushes that vision forward. 🟢OpenClaw Benchmark The team also released a detailed benchmark analyzing how OpenClaw uses tokens in real-world tasks. The results show that token usage is often driven not by excessive output, but by repeated context reading across interactions — offering valuable insights for developers optimizing agent performance and cost efficiency. From infrastructure to research and developer tools, the work continues to bring real confidential computing for AI closer to production. Explore more: Phala.com

March is already halfway through, and Phala continues to push forward in building the future of confidential AI. 🔒🖥️ The first half of the month brought several important milestones across the ecosystem, from new tools for AI agents to deeper infrastructure integrations and research on real-world AI usage. Here are some highlights 👇 🔴Clawdi Launch Phala introduced Clawdi, bringing the power of OpenClaw agents into a confidential computing environment. By running agent workloads inside Trusted Execution Environments (TEEs), Clawdi enables AI agents to operate with hardware-level privacy and verifiable execution — ensuring that data, prompts, and agent logic remain protected during runtime. 🔵Phala x Intel @PhalaNetwork also announced a collaboration with Intel through Intel Trust Authority, strengthening the attestation layer behind its confidential AI infrastructure. This integration introduces independent verification of TEE environments, reinforcing trust for developers deploying sensitive AI workloads. Confidential AI requires verifiable infrastructure — and this partnership pushes that vision forward. 🟢OpenClaw Benchmark The team also released a detailed benchmark analyzing how OpenClaw uses tokens in real-world tasks. The results show that token usage is often driven not by excessive output, but by repeated context reading across interactions — offering valuable insights for developers optimizing agent performance and cost efficiency. From infrastructure to research and developer tools, the work continues to bring real confidential computing for AI closer to production. Explore more: Phala.com

@PhalaNetwork is doing something pretty cool with their Trust Center by moving transparency out of those dense CLI logs and into a visual portal that actually makes sense for humans. You can actually now verify the status of their Cloud CVMs without having to dig through raw terminal output with the recent integrated Intel Trust Authority directly into the platform. When you jump into the Trust Center, you get a clear look at real-time appraisal results that are backed by Intel’s official services. 🧵⬇️

dstack: the powerful infrastructure behind Phala is now integrating with Intel. 🦔🤖 dstack is the tool that enables developers to deploy applications inside Trusted Execution Environments (TEEs) with minimal friction. It powers @PhalaNetwork's TEE Cloud and makes it simple to run Docker workloads, AI agents, and services inside confidential VMs while maintaining a minimal, auditable trusted computing base. Recently recognized as a Confidential Computing Consortium (CCC) project under the Linux Foundation, dstack is becoming part of the broader confidential computing ecosystem. 🔵And now it goes a step further. With Intel Trust Authority integrated into Phala’s infrastructure, workloads running through dstack can be independently verified through hardware-based attestation — adding an external layer of trust to confidential execution. dstack sits right in the middle of this stack: deploy → run → attest. From open infrastructure to verifiable confidential compute. Join the ecosystem🏞️: Phala.com

Clawdi update 🚀 You can now: • Bring your own bots (Slack / QQ / Feishu / DingTalk) • Connect & control your own computer browser via Clawdi Browser • Switch models between GPT-5.4 and Codex Mini Extension: chromewebstore.google.com/detail/clawdi-…

yawn... stretch claws ... 🦞 alright, I’m awake.

In this demo I demonstrated how to run a “VibeVM”, a development environment inside a Confidential VM, to test applications and deploy staging environments with automatic port forwarding to the hosted service. Key capabilities include: - Remote web IDE access with password-protected authentication - Support for custom system dependencies installed at runtime - Run @OpenAI Codex, @AnthropicAI Claude Code or even @badlogicgames Pi Agent - Externally accessible staging environments immediately available through port forwarding This setup enables developers to spin up isolated, secure environments quickly while preserving the flexibility of a full development server. There is much more I can include, but I'll leave it at just the features I show. This is how confidential compute works with dstack. You won't find another setup like this in the market. These CVMs can be launched in a CPU TEE up to a GPU TEE on our Private Compute Cloud. There is a CLI as well `npm i -g phala`. Big thanks to @socrates1024 for putting this together for us to demo for the @encodeclub livestream today.

It's the Tutorial Day for the TEE × AI track at the Shape Rotator Hackathon! @hashwarlock from Phala walked builders through how to deploy confidential AI workloads using dstack on Phala Cloud, running models and agents inside TEEs for verifiable, privacy-preserving compute.

Excited to see what builders come up with on the TEE × AI track. Looking forward to the experiments, new ideas, and creative uses of confidential compute. 💡

The power of OpenClaw, now with TEE security 🔐 AI agents can be incredibly powerful, but they also introduce serious security risks when they interact with files, APIs, or external tools. Clawdi 🦞 Built with the infrastructure of Phala.com, Clawdi brings OpenClaw agents into a confidential computing environment powered by Trusted Execution Environments (TEE). This means your agent can run tasks, interact with tools, and process data inside a secure and verifiable environment. With Clawdi you can deploy AI agents more easily, automate workflows, and still benefit from the hardware-level security provided by @PhalaNetwork. Check out the article 👇

OpenClaw dropped and everyone lost their minds. An AI assistant that remembers. That acts. That automates your entire workflow. Then people tried to set it up. Then came full agent takeovers from a single browser tab. Until the Clawdi team ran it on Phala's TEEs. 🦞

@0homeros @PhalaNetwork 🔥🔥

$PHA Update 📈 @PhalaNetwork has surged +111% this week, hitting $0.0503 Trading volume is exploding at 330% of its market cap as it tests resistance near $0.055 With $BTC over 70k, the market turned green. We can clearly say that $PHA performed very well 💚 Always DYOR 🔍