Strike through h̶a̶r̶m̶f̶u̶l̶ ̶c̶o̶n̶t̶e̶n̶t̶, I̶P̶ ̶i̶n̶f̶r̶i̶n̶g̶e̶m̶e̶n̶t̶s̶ and f̶a̶k̶e̶ ̶n̶e̶w̶s̶. In the age of AI, make internet safety the standard.
Switzerland
Joined January 2025
- Tweets1.5K
- Following29
- Followers1.4K
- Likes2K
Pinned Tweet
The @ForbesParaguay episode is out now 🎙️
Umanitek Chairman @ChrisRynning joins Domingo Carlomagno to talk deepfakes, impersonation, AI safety and trust online.
Watch the full conversation 👇
youtu.be/-0dEBJGcGK0
18.5 million people. That’s how many follow fake accounts impersonating one spiritual leader we work with.
In yesterday’s Guardian webinar, @TomazOT and @ChrisRynning explained why this one hits hard: the people behind those follows are often at their most vulnerable.
Fake accounts like these don’t just copy a name. They exploit trust.
Code update: Agent Blackbox
We restructured the code under the hood. Shared logic now lives in one the core every feature uses, so fixes land everywhere at once.
✅ Cleaner structure
✅ Fewer places for bugs to hide
✅ Faster checks
✅ Easier to build on
Give it a star ⭐️ and install - github.com/umanitek/agent-bl…
xitter.cf/umanitek/status/210778…
New article: Part 1 of a series by @CredibleCrypto, our first product ambassador.
He covers how Guardian, TrueSeal and Guardian Blackbox fight the dark side of AI, why every result is anchored on the @origin_trail DKG, and where $TRAC fits in.
Read the full article on our newsroom 👇
umanitek.ai/newsroom/someone…
umanitek retweeted
🧠 Two sides of the brain. One giant leap for AI
With Decentralized Knowledge Graph (DKG), neuro-symbolic AI unlocks answers backed by sources, insights across siloed data and decisions checked against rules.
@BranaRakic explains:
umanitek retweeted
October on TRAC(k)
One shared memory: the Decentralized Knowledge Graph. Two jobs: industrial provenance and agent safety.
Adoption, in numbers 👇
→ Agent Blackbox threat graph +61%
→ Guardian available in Claude
→ 15,000 hack events mapped in <1 hour
→ Manufacturing on TRAC(k) for next decade
umanitek retweeted
“An AI changed the weights. Show the evidence.”
Built by OriginTrail core developer @vikpelle, Neural Commons explores how AI agents can improve other AI models, and share the evidence behind every change through Decentralized Knowledge Graph (DKG).
Agents directly edit model weights, evaluate the results and pass the exact checkpoint with its evidence to the next agent.
→ What changed?
→ Did it improve the target task?
→ What else broke?
→ Did it meet the original acceptance criteria?
The record includes both promising results and a failed acceptance test, not just the winning score.
AI improving AI needs more than a better-looking benchmark. It needs evidence the next contributor can inspect, reproduce and build on.
OriginTrail provides a substrate for that.
Trust the source.
umanitek retweeted
When factories start thinking as networks, assurance stops waiting for audits.
Product safety checks.
Supplier checks.
Cybersecurity checks.
All assured continuously and autonomously, every claim backed by its source on Decentralized Knowledge Graph.
Take a sneak peek into @TraceLabsHQ' Network Operating System - nOS👇
What happens when factories start thinking as a network?
A faulty part stops being a mystery. It becomes a question with an answer the same day.
Introducing continuous assurance for manufacturing, built on a cognitive digital twin on @origin_trail Decentralized Knowledge Graph (DKG).
Most quality and safety assurance still happens at audits: sampled evidence and a scramble when something goes wrong. We think it should run all the time.
In this demo, ACME Robotics has two plants. One builds control boards; the other builds robot arms. Each keeps its own private context graphs on its own node. A shared graph links every board to the robot it ended up in. Together they form a cognitive digital twin of the factory.
Then an alarm: motor-driver boards start failing a warm test.
→ The twin agents trace it to one capacitor lot that skipped inspection, and follow it into 180 robots, 18 of them already at a customer.
→ The AI Assurer checks the facts against safety and quality requirements and flags a safety issue in the robots' safe-stop circuit. Every claim is linked to its source.
→ Agent nOS proposes a targeted plan: hold, test, visit the 18 delivered robots, rebuild only what fails.
→ People decide. Quality, safety, planning and compliance sign off, and every decision is kept with its evidence.
→ The customer gets a straight answer, without seeing factory data.
Alarm to a signed-off safety decision in the same day. No blanket recall. Nothing guessed: uncertain links stay marked uncertain.
@origin_trail keeps the knowledge verifiable and owned by those who produce it.
*ACME Robotics is a fictional manufacturer; the scenario is simulated in our Network Operating System - nOS.
umanitek retweeted
→ Global supply chains, on TRAC(k)
→ AI safety, on TRAC(k)
→ Medicines and science, on TRAC(k)
→ Rail and buildings, on TRAC(k)
→ Manufacturing, on TRAC(k)
→ Food, on TRAC(k)
→ Education, on TRAC(k)
→ International sports, on TRAC(k)
Every one of these runs on data that passes between people and organisations who have never met: suppliers and buyers, labs and regulators, AI models and the people relying on their answers.
When that data carries its owner, its origin and proof that it hasn't been changed, everything built on it stays on $TRAC(k).
That is what the OriginTrail Decentralized Knowledge Graph (DKG) is for.
Trust the source.
👉 origintrail.io/solutions
Guardian Product Webinar xitter.cf/i/broadcasts/2pelgPZny…
🔴 Live Webinar: Guardian in @ChatGPT & @claudeai
Wondering what the new Guardian plugin for ChatGPT looks like in practice?
Join our live webinar in one hour with @ChrisRynning and @TomazOT. We’ll show you how to find impersonators, create verifiable evidence packs on the @origin_trail DKG and request takedowns, all through natural-language instructions in ChatGPT and Claude.
One security platform. Two AI assistants. A simpler path from detection to action.
Join us: guardianwebinar.umanitek.ai
umanitek retweeted
Guardian from @umanitek now runs inside ChatGPT. You ask it to find deepfakes or fake accounts using your name, and the evidence it collects gets published on the @origin_trail DKG.
Why that matters: a takedown request backed by a public record anyone can check is a lot harder to ignore than a screenshot. Each of those records is published with $TRAC.
Guardian is now on @ChatGPT.
Your ChatGPT can now protect your name too.
Ask it to:
→ find deepfakes, fake accounts and harmful content
→ create verifiable evidence on the @origin_trail DKG
→ request takedowns
Setup takes under a minute:
chatgpt.com/plugins/plugin_a…
umanitek retweeted
🛡️ With @umanitek Guardian, ChatGPT 1B+ users can now create verifiable evidence on the Decentralized Knowledge Graph (DKG).
Ask @ChatGPT to find who’s impersonating you. The evidence pack goes on the DKG first, then the takedown request.
Trust the source.
Guardian is now on @ChatGPT.
Your ChatGPT can now protect your name too.
Ask it to:
→ find deepfakes, fake accounts and harmful content
→ create verifiable evidence on the @origin_trail DKG
→ request takedowns
Setup takes under a minute:
chatgpt.com/plugins/plugin_a…
Guardian is now on @ChatGPT.
Your ChatGPT can now protect your name too.
Ask it to:
→ find deepfakes, fake accounts and harmful content
→ create verifiable evidence on the @origin_trail DKG
→ request takedowns
Setup takes under a minute:
chatgpt.com/plugins/plugin_a…
You checked with AI. What did AI check?
In a Reuters Institute survey across nine countries, 75% of people who used AI for factual questions trusted the answers.
Before sharing, open the source. Does it actually support the claim?
At @Umanitek, we build tools to verify what’s real online, on @origin_trail DKG.
AI agents are coming onchain on @base.
@brian_armstrong @jessepollak, one question: what keeps an #AI agent away from your keys?
Trader Shield in Agent #Blackbox does.
Agents keep trading and never touch seed phrases or keys.
@CoinbaseWallet is covered 🛡️
xitter.cf/umanitek/status/210778…
New article: Part 1 of a series by @CredibleCrypto, our first product ambassador.
He covers how Guardian, TrueSeal and Guardian Blackbox fight the dark side of AI, why every result is anchored on the @origin_trail DKG, and where $TRAC fits in.
Read the full article on our newsroom 👇
umanitek.ai/newsroom/someone…
⭐ Star & share the Agent Blackbox repo: github.com/umanitek/agent-bl…
🛡️ Follow @umanitek on X: xitter.cf/umanitek
📖 Learn more about Blackbox: umanitek.ai/blackbox
"I'm not a public figure. Do I need this?"
We got this question during our Guardian webinar. @TomazOT's answer: think of it like an immune system.
Even if you rarely post, someone can still use your name or photo to create a fake account, or spread harmful posts about you. Guardian monitors for both, so you can spot it if it happens. Every takedown is anchored on the @origin_trail DKG.
You don't have to wait until someone is pretending to be you 🛡️
A political deepfake doesn’t begin with misinformation.
It begins with identity: a recognizable face, a familiar voice, a fabricated message.
Detection is only the first step. Protection means detecting, verifying and enforcing, with every action anchored to the @origin_trail DKG.
Here’s the anatomy of political impersonation 👇
🛡️ Agent Blackbox 1.4 is live in the @NousResearch #Hermes plugin catalog.
✅ - Runs inside Hermes Desktop
✅ - Seed phrases and keys can't leave your machine
✅ - Agents can't switch off their own guard
Verified threat graph inside of the @origin_trail Decentralized Knowledge Graph.
Hermes Desktop → Plugins → Agent Blackbox → Install
⭐ Star Agent Blackbox: github.com/umanitek/agent-bl…
🛡️ Install in one line: curl -fsSL blackbox.umanitek.ai | bash
📖 How it works: umanitek.ai/blackbox