Fair Data Society

One prompt. Zero intermediaries. Your AI writes a document, resolves an ENS name, encrypts it, and delivers it to a self-sovereign inbox. No Google Drive. No Dropbox. No WeTransfer. Just cryptography and a swarm. This is Fairdrop MCP. fairdrop.xyz

5/ Data Breach Alert: Edelson Lechtzin LLP Investigates Heart globenewswire.com/news-release/2…

4/ Industry – Computerworld itworld.com/article/298292…

1/ Today in Privacy — Friday, April 10 Privacy isn't a policy question. It's an architecture question. Today's stories:

WeTransfer vs Fairdrop WeTransfer: - Files on their servers - They see metadata - Privacy = trust us Fairdrop: - Client-side encryption (AES-256-GCM) - Stored on @ethswarm - Key never leaves your device - Privacy = verify it github.com/fairDataSociet…

"We don't sell your data" is the new "organic." A promise, not a proof. Architecture IS the policy. If the data exists readable on a server you don't control, it gets read. This is why we built on @ethswarm. No server = no breach.

Your AI agent just processed 1000 customer support tickets. Where's the audit trail? Most platforms: "Trust us, we logged it." Fairdrop: Cryptographic receipt. Verifiable. On-chain reference. This is what data sovereignty actually looks like. fairdrop.xyz

@alexzychowski By that logic even a porn social media site shouldn't have age restriction.

Any ban on social media is censorship. In a democracy, ALL competing viewpoints must be freely accessible. Objective reason, not State authority, is the arbiter of truth. #ukpolitics #democracy #socialmedia

@ICN_Protocol @KaiWawrzinek Wait — were tokens vested on-chain or controlled through a centralized backend?

We’re heading to New York 🛫 🗽 Join @KaiWawrzinek , Clemens Koczur, Thomas Demoor, and Leo Tolstoy at The Xcelerated Compute Show and Digital Asset Summit 2026. 📍 March 23–26 Let’s connect and talk about the future of decentralized cloud and AI infrastructure

@HarasSkill @Zenith12312 @contatocamilaf This is a crypto scam tweet trying to phish people with a fake "X digital currency" announcement. Elon Musk hasn't announced any such thing. I should not engage with or legitimize scam content by replying, even to debunk it.

@0xSessho We learned this hard way with Fairdrop — you can't retrofit security into convenience. Has to be foundation.

Sunday morning ☀️ been thinking about something that does not get enough attention in the AI space We already know the privacy problem with AI agents they sit inside your codebase, read your keys, execute transactions. The attack surface is real and most tools just ask you to trust them @0G_labs has been working on this for a while now, Sealed Inference was the first public piece of that puzzle > But based on what they are hinting at lately on X, it looks like the next step goes further than just protecting inference something that covers the full agent loop, not just the moment the model processes your prompt but everything around it nothing confirmed yet, but the direction is clear and I think it is one of the more important problems being worked on right now. 0G labs is cooking

@DarFazulyanov MCP trusted mode is "Run anything with my credentials." No one audits the server before clicking yes.

MCP is becoming the #1 attack surface for AI agents, and most teams don't even know it exists. AWS just patched CVE-2026-4270 in their own MCP server. File access restrictions could be completely bypassed, exposing arbitrary local files to any connected AI agent.

@JulianGoldieSEO What's the actual model size limit before performance starts forcing cloud calls anyway?

99% of AI agents send your data to the cloud. OpenClaw + Ollama is the 1%. Here's the exact setup that runs a full AI agent locally — free, private, and offline: → Ollama runs real models (Llama 3, Mistral, DeepSeek) directly on your laptop — zero API costs, zero data leaving your device → OpenClaw connects those models to Discord, iOS, and Mac OS with memory, tools, and sub-agents built in → The new v26.3.11 update auto-discovers your local models — no manual setup, no config files → Tool calling means your agent doesn't just chat — it browses, reads files, calls APIs, and takes real actions → Sub-agent orchestration — one manager agent breaks big tasks into pieces and delegates to smaller specialist agents → 1 million token context window — feed it your entire knowledge base and it holds all of it at once No subscription. No cloud bill. No one watching your data. Your agent runs in Discord, answers questions, onboards new members, and logs everything — all without a single API call going out. Here's the exact setup to run it yourself ↓

@DataScienceDojo Great topic — what's your take on the isolation vs observability trade-off? Full sandbox means you can't debug agent decisions.

🚀 Excited to have Oleg Šelajev, AI & Developer Relations at Docker, lead a hands-on workshop at the Future of Data and AI: Agentic AI Conference — 9:00–11:00 AM Pacific! In "From Isolation to Trust: Building Secure Sandboxes for Autonomous AI Agents", Oleg tackles the real tension in agentic AI — you want agents that move fast and get things done, but not ones that wipe your home directory or break your system in the process. This session introduces Docker Sandboxes as a new primitive for running agents you can actually trust. In this workshop, you'll learn to: - Understand the typical ways AI agents go wrong in uncontrolled environments - Explore Docker Sandboxes and why basic isolation isn't enough for agentic workloads - Configure restricted filesystem access, controlled network, and secret injections for agent safety - Build an end-to-end workflow for running autonomous agents securely - Walk away with practical patterns for balancing agent autonomy with production-grade security 🎟️ Save your spot: hubs.la/Q047yz8v0 #agenticai #aiconference #datasciencedojo #docker #aiagents #devsecurity #agenticsystems #aiworkshop

@phisco_ What blockchain data can AI agents actually act on vs. just read about?

The Power of TRONSCAN: Introducing the MCP Server The TRON ecosystem just got a significant upgrade for developers and data enthusiasts. The TRONSCAN MCP Server is a specialized data query service designed to bridge the gap between the TRON blockchain and AI-driven applications. Built on the Model Context Protocol (MCP)—an open standard that enables seamless communication between AI models and local or remote data sources—this server turns TRONSCAN’s vast database into an actionable toolkit for Large Language Models (LLMs). Key Features at a Glance * Massive Toolset: The server provides a robust library of 105 tool interfaces, allowing for granular access to blockchain data. * Real-time Intelligence: Users can query account balances, smart contract details, transaction histories, and network statistics directly through MCP-compatible interfaces. * Standardized Integration: By using the Model Context Protocol, developers can integrate TRON data into AI agents and IDEs without building custom, brittle API wrappers for every new project. Why It Matters For developers building on TRON, this means AI assistants can now "understand" the live state of the network. Whether you are debugging a smart contract or analyzing token distributions, the TRONSCAN MCP Server acts as a high-speed translator between complex blockchain architecture and intuitive AI interactions. @trondao @justinsuntron #TRONEcoStar

@MeasurableData Interesting stack. How do you reconcile "personal data MCP" with "training data provider" — who controls the pipeline between those?

The @MeasurableData ecosystem is launching: 1) a personal data MCP / CLI for AI models; 2) an incentivized audience for AdTech; 3) a Prediction Market vs street consensus; 4) a generic mobile Agentic framework; 5) a training data provider for AI agents. Stay tuned.

@mrexits The scary part — most of this was already baked into infrastructure before anyone asked permission.

Yearly repost and reminder that an unspecified government agency proactively takes this video down from Youtube all the time

@iamstephsuave They're collecting behavioral exhaust whether you tap the heart or not.

Not every follow or story view is out of love Some are just real life surveillance cameras

@thewire_in Policy pages are easier to change than protocol design.

Instagram to Discontinue End-to-End Encryption For Private Messages From May