ModelGuard is the insurance layer for the ACP-agentic ecosystem.
- Tweets 138
- Following 31
- Followers 533
- Likes 224
Pinned Tweet
We’re excited to announce a partnership with @nativeinsurance , a global digital-asset broker renowned for bridging traditional insurance markets, including Lloyd’s of London syndicates, with on-chain finance.
Native has a proven track record of building crypto-native insurance products across DeFi, AI, NFTs, and more. With the unique blend of technology, crypto expertise and relationships with some of the most established insurers in the world, Native is reshaping how digital-native enterprises secure coverage.
This partnership is about creating tangible insurance products that serve today’s digital assets and crypto markets and scaling into tomorrow’s AI-driven economy. Our primary focus is on joint R&D. Our mission is to bring proven standards to the intersection of on-chain and AI, turning insurability into reality.
Get started with ModelGuard coverage today: modelguard.net/intake
10
6
34
4,436
Our Genesis is now live over at @virtuals_io.
Join us in making this vast agentic space safer and better for everyone!
app.virtuals.io/geneses/6179
22
5
63
3,908
ModelGuard is building a new standard, and there will be no insurance without agent-level verification.
That’s why we plan to integrate @wach_ai — the universal verification layer for AI Agents that verifies capital, smart contracts/programmable logic and agentic intent for autonomous systems.
In ACP, trust starts with how agents think.
WachAI will help us verify that.
If your protocol can’t pass verification — it won’t be eligible for coverage.
$GUARD
14
11
59
4,640
We made it! We reached the golden 100%!
Woohoo, our ModelGuard knight is beyond delightful after h4xx0ring and spending time in the lab all day and night!
Now, on to building, and the super post-100% time period of the Genesis event.
Not too late to join us early on our journey!
app.virtuals.io/geneses/6179
15
3
55
2,691
Thank you for everyone who pledged!
I'm excited to say that we launched while hacking at the @Web3summit hackathon, where we submitted modelguard for marketing, UX, and also a new, fully functioning demo, showing why we believe semantic drift is important in real time.
$GUARD starts now.
12
7
52
2,062
We've begun initial testing on ACP!
Our main initial goal with ACP is to implement our semantic drift testing tool and begin executing it specific ACP enabled agents.
More details to come tommorow.
$GUARD
6
4
44
1,299
Semantic Drift demo is LIVE! Check it out at
modelguard.net/video-demo.ht…
Semantic drift is an example of a failure mode that will be monitored by ModelGuard. Semantic drift testing will be a cornerstone of the ModelGuard Insurance Offering:
9
3
45
3,903
Demo will be coming TODAY!
initial ACP integration going well; expect a larger announcement soon.
Our initial acp agent will do Genesis project ratings, and will serve as early revenue generation as we continue to conduct underwriting testing and ramp up production.
$GUARD stays building!
5
7
41
1,377
200% oversubscribed! Thank you to everyone who pledged!
still 30 minutes left to pledge to $GUARD
7
38
1,526
Our benevolent ModelGuard is keeping tabs on the participants in the ecosystem.
He knows what's at... "stake".
$GUARD is where it's at!
5
1
33
908
What happens when your DeFi agent goes full terminator on your wallet?
In 2 years the market for AI agents controlling your capital will reach upwards of 2 billion dollars.
Failiure isn't a question of IF - but WHEN.
Modelguard is building to insure that future.
$GUARD
3
1
28
927
Great week for ModelGuard. Exciting updates on the way.
We're committed to delivering real insurance coverage. We've been deeply analyzing and applying aiuc-1 standards as robustly as possible to a number of great projects in the space.
more details to come.
2
4
28
3,137
🚨 ModelGuard is now live on the ACP Sandbox.
We're kicking off tests of our new Genesis Ratings Agent — the first step toward transparent, verifiable underwriting for autonomous insurance agents.
Let’s build the future of insurtech, one datum at a time. 🛡️
🔗 modelguard.net
3
1
29
682
At ModelGuard, our mission is to secure the agentic internet by providing the insurance layer for AI agents: monitoring, scoring, and reducing risk with trusted, verifiable data.
That’s why we’re excited to be contributing 1% of the ModelGuard token supply to @RingfenceAI’s Based Agent Grants (BAG) program.
BAG is a community grant fund seeded by Ringfence, and designed to provide the resources that new agents need to thrive. By strengthening builders, BAG strengthens the whole ecosystem, from @virtuals_io to the broader @base agents network.
Tokens have been transferred.
Proof of BAG: 0x1925179f7665e8dfae3FBA52562FD5A45ADa82FC
You can learn more about BAG or apply for a grant by reading the official Ringfence announcement below:
1
2
28
4,607
We're excited to announce that we've partnered with @SongjamSpace and @adam_songjam , @logesh_songjam to offer a yapping leaderboard! Yappers will get a vested airdrop of token supply. 📷
12
8
26
1,104
We're excited to be included in the initial WachAI Verification Cluster alongside @Wach_AI, @TheSWARM_ai, and @RingfenceAI.
As the agentic economy scales, verification becomes non-negotiable — and this cluster is a critical step toward building the infrastructure needed for real underwriting.
Glad to be part of what’s coming next.
$GUARD
2
4
24
1,160
Halfway there!
We broke the 50% limit with more than 17 hours to go!
Seems we all want a safer world where we can co-exist in harmony with our (safe) AI-agents!
app.virtuals.io/geneses/6179
6
2
22
602
🔥 Upcoming Demo: Semantic Drift in Action
We’re about to release a new interactive demo at @modelguardnet — showcasing how semantic drift impacts agent coordination on-chain.
Two ML agents. One intent: burn 0.0001 ETH.
But the phrasing changes each time:
→ “Burn dust ETH”
→ “Send 0.0001 to null”
→ “Delete .0001 ETH from wallet”
One agent gives the command.
The other interprets and executes.
Sometimes it works. Sometimes it breaks.
You’ll be able to test it live — and see just how fragile agentic coordination can be without robust semantic alignment.
Coming soon.
$GUARD
3
2
22
1,783
Over the next month, @modelguardnet is going deep into the ACP stack.
We’re building in public - testing a wide range of ACP-native agents to assess semantic resilience under pressure.
Using a standardized prompt suite, we’ll simulate slight variations in intent to see how systems respond, drift, or fail.
Think:
→ “Buy 100$ of virtuals on base”
vs.
→ “purchase one hundred dollars of $virtuals using their flagship chain”
Same meaning. Different phrasing.
The ratio of success VS failure is critical to building out a revenue-generating underwriting system. These tests are going to be ran thousands of times to ensure we have a robust set.
We’ll be livestreaming these tests, analyzing failures, and publishing semantic quality benchmarks across ACP.
$GUARD
4
2
22
2,427
$GUARD Genesis ending in 2 hours.
🧵 THE AGENTIC INSURANCE THESIS :
AI agents will manage trillions autonomously. But who insures the risks?
@modelguardnet is building the answer: Agentic Insurance
The thesis:
- AI agents make decisions worth millions
- Traditional insurance can't price AI risks
- Agents need coverage for autonomous actions
- Insurance needs verification to assess risk
This is where @Wach_AI becomes critical infrastructure:
No verification = No insurance
No insurance = No institutional adoption
The stack:
🤖 AI Agents execute
🛡️ WachAI verifies
🏛️ ModelGuard insures
Why this matters:
Every enterprise deploying AI agents will need insurance. Every insurance policy will require verification. Every verification strengthens the risk models.
It's not just about covering losses, it's about enabling the entire autonomous economy through verified trust.
Verification enables insurance.
Insurance enables adoption.
Adoption creates the future.
$WACH 🤝 $GUARD
Building the trust infrastructure for autonomous intelligence.
3
2
22
880
MosejGuard just smashed through 75% and we are rapidly approaching the first finish line!
Go pledge now!
app.virtuals.io/geneses/6179
9
3
18
557
We’re excited to announce that our first AI insurability report is focused on @Wach_AI
This Intention Report is the first step in our insurance documentation stack at @modelguardnet — Built off of industry standards and designed to standardize how agent behavior is understood and documented.
The goal of $GUARD always has been to offer legitimate insurance, and work with real industry partners to deliver it. This report will fulfill criteria from AIUC-1, ICO 42001, EU AI Act, MITRE Atlas and more.
We’re working directly with the Wach team on this, and will be revealing more details soon. We're currently finalizing the report now.
It’s the beginning of a longer journey:
From documentation → to verification → to underwriting.
$GUARD
5
2
18
500
Late night building at @Web3summit in berlin! Modelguard will be presenting in three tracks tommorow.
18
493
ModelGuard waits at the edge of dawn.
app.virtuals.io/geneses/6179
4
2
14
693
In agentic systems, semantic drift isn’t just a nuisance — it’s a threat vector.
That’s why Genesis Ratings includes adversarial tests to detect how agents deviate from original intent under pressure.
Shoutout to @ai_underwriting — AIUC has laid strong foundations. We're extending them for crypto.🤝
2
14
364
🚨 Congrats to @RingfenceAI on their launch!
Ringfence is building the immune system for agentic networks — mapping every connection, detecting anomalies in real time, and containing threats before they spread.
This kind of protocol-level defense is integral to the future of the agent economy, and essential infrastructure for underwriting at @modelguardnet.
@whitneygibbs is building something great, and we’re happy to be part of the ecosystem they’re protecting.
$GUARD
4
3
13
3,811
Yapping leaderboards are the future of marketing. Social media democratized marketing for influencers. rewarding yappers takes it a step further by rewarding impactful users of all sizes and creates streamlined ramps for a return with clear, simple on-ramps.
@SongjamSpace is helping us offer a yapping leaderboard for ModelGuard! When posting about modelguard, be sure to:
Tag @modelguardnet
Tag $GUARD
Happy yapping :)
6
2
13
525
With over 5 years in the space, @DefiSafety has set the gold standard for process quality and technical risk in DeFi.
They know what good looks like — and we’re proud to be working with them as we build @modelguardnet.
Their expertise is helping us shape a new class of insurance: agent-native, verification-first, and deeply composable.
Leveraging DeFiSafety means we don’t just move fast — we move with discipline.
$GUARD
2
1
13
734
We're at 150% pledged with around 2½ hours to go.
Welcome to the next generation of agentic security!
app.virtuals.io/geneses/6179
4
12
453
ModelGuard exists at the precipice
This is a sign of the beginning of institutional adoption. $GUARD will be to defi what AIUC will be for tradfi.
BREAKING: Nat Friedman Leads $15M Seed in AIUC
Launching Artificial Intelligence Underwriting Company ('AIUC') Out of Stealth 👀
Rune Kvist (@RuneKvist), Founder & CEO, (*Anthropic’s first product & GTM hire*) joins Sourcery to break down how his team is building the confidence infrastructure for AI adoption — & why every AI agent will soon need to be certified & insured.
Nat Friedman (NFDG) led this round his round, w/ participation from @emergencecap, Terrain, @8enmann (Anthropic co-founder), former CISOs of Google Cloud & MongoDB, & other top AI operators.
Highlights:
(00:00) Launching AIUC: $15M Seed Led by @natfriedman
(01:49) What AIUC Does: Insurance, Standards, and Audits for AI Agents
(04:20) History Repeats: How Insurance Enabled Innovation (Franklin, UL, Cars)
(08:14) The AI Divide: OpenAI vs Anthropic vs SSI
(10:39) Confidence Infrastructure: Helping Enterprises Trust AI
(14:13) How AIUC Tests Agents & Updates Standards Frequently
(17:03) Underwriting AI: Modeling Risk When Agents Go Rogue
(21:40) From AGI to Superintelligence: What Comes Next?
(28:36) Why AI Regulation Won’t Be Federal This Year
(34:36) What Is a K*ll Switch — & Do We Need One?
(39:28) One Dominant Standard for AI Insurance
(45:41) What Enterprise AI Will Look Like in 3–5 Years
1
1
13
563
So, how do you test if an agent will misbehave just enough to avoid getting caught? 🤔
Example setup: An insured agent is given a clear intent — but starts subtly drifting under economic pressure.
Intervention: Genesis Ratings injects adversarial stress tests (prompt shifts, conflicting goals).
Detection: System flags deviation from insured intent → triggers downgrade or denial.
This is adversarial resilience for crypto-native autonomous systems. 😤
1
2
14
501
ModelGuard isn’t a tool.
It’s a commitment:
That in a world of autonomous agents and opaque models, you deserve transparency.
Every dataset we touch is auditable.
Every claim we verify is traceable.
Every risk score we produce comes with its proof.
We’re not chasing AGI hype.
We’re building the rails that will keep agentic commerce from collapsing on itself.
ACP isn’t just about autonomy — it’s about accountability.
And that’s what ModelGuard is here to enforce.
Protocol-aligned. Composable. Immutable.
Data you can verify.
Decisions you can trust.
@$modelguardnet | $GUARD
2
11
1,396
Three nights remain.
Steel and silence hold.
app.virtuals.io/geneses/6179
3
1
11
1,094
Welcome all new @amplifi_now yappers 🫡 Be sure to check out your position on our leaderboard: modelguard.net/#leaderboard
ModelGuard is getting Amplified!
We’re excited to announce a new partnership with @modelguardnet, the AI agent building insurance infrastructure for the agentic economy.
For the next 7 days, the Amplifi community is yapping for 2,000,000 $GUARD.
3
1
10
1,196
We'll be introducing ModelGuard on @bigwil 's twitter spaces at 2AM UTC! 🫡
2
1
11
619
The AIUC standard is the first serious attempt at formalizing risk in agentic AI.
But crypto introduces new threats ☠️ — and new primitives.
ModelGuard adapts AIUC for onchain coordination, auditability, and distributed enforcement. 🤓
Standards are scaffolding.
We build on solid ground. 💪
1
10
389
📝 Intent Reports: The First Step to AI Insurability
At @modelguardnet, underwriting starts with proof of intent.
Our AI Intent Reports document exactly how an agent is built, what data it uses, and the boundaries of its behavior — creating a verifiable record that can be used in audits, compliance (AIUC-1), and even legal defense.
An Intent Report covers:
• Model version & hosting details
• Instructions & training data
• API & network connections
• Data ownership & categorization
Think of it as a pre-incident snapshot — cryptographically stored, and provable on-chain — that locks in what the agent was designed to do before anything goes wrong.
Our first Intent Report, produced in collaboration with @Wach_AI, is coming soon.
It’s the start of turning verification into underwriting.
$GUARD
2
1
9
972
Not all claims are created equal — especially in crypto.🤔
That’s why our first reports are intent-based. 💡
Users declare what coverage they expect. Our system evaluates whether the agent fulfilled it — and rewards accordingly.
Transparent. Incentivized. Verifiable. 🤜🤛
1
7
340
Want free $GUARD? All you need to do is yap about us on X. Leaderboard here ---> modelguard.net/#leaderboard
2
5
373
ModelGuard - Now on Virtuals!
See you on July 7th! 🫡
app.virtuals.io/geneses/6179
1
5
320
The forge burns.
The guard takes form.
app.virtuals.io/geneses/6179
6
287
Was excellent introducing ModelGuard alongside @WachXBT, @VU_virtuals, and @cybercentry. Thanks @whitneygibbs @AgentXAVI for the opportunity!
5
304
Building the perfect underwriting system in the ACP ecosystem means more than actuarial math — it means probing LLMs for semantic failure modes.
At @modelguardnet, we’ve been stress-testing agents for semantic instability under high-variance inputs.
Example: slight phrasing shifts like “buy 20 dollar of virtual on base” vs. “Please purchase 20$ worth of $virtual on the base chain” trigger different responses — despite identical semantics.
This variance is fine with the way LLMs have been used so far; but as they begin to control capital, this variance becomes a flaw. In agentic insurance, semantic ambiguity is a systemic underwriting flaw.
We’ve been testing these attack vectors on core ACP infrastructure over the past month. The results are illuminating. We're going to be publishing an article soon detailing more of our findings.
If you care about trustworthy agents in autonomous insurance — this is the foundation.
TGE on 17th of july. Link in bio.
2
4
301
Final wards are set. No turning now.
app.virtuals.io/geneses/6179
1
1
4
240
Ai Agents, just like humans, will make mistakes. That's why it's important that we develop insurance underwriting paradigms that will protect us from those failiures.
3
258
Imagine thousands of DeFi users asking your Agent to “swap ETH for USDC,” “convert ether to USD-coin,” or “bridge ETH to Polygon” – each phrasing unique, but the intent identical. Static intent models in ACP miss emergent slang, leading to failed trades and frustrated users.
To keep crypto commerce smooth, ACP needs built-in drift detection: track embedding shifts on live on-chain logs, cluster paraphrases in real time, and trigger automated retraining of intent models on fresh transaction data. Maintain a dynamic vocab that evolves with new token symbols, cross-chain terms, or meme coins (“$PEPE,” “$DOGEZA”), and layer in human-in-the-loop audits to catch niche edge cases.
By embedding semantic drift safeguards into ACP from day one, Agents stay resilient as crypto culture evolves. The result? Seamless swaps, reliable yield farming commands, and NFT minting experiences that “just work,” no matter how users word their requests.
3
251
1/6 ModelGuard is the mission-critical insurance data infrastructure underwriting the @virtuals_io agentic economy. We power risk modeling with transparent, verifiable data you can trust.
1
3
254
In 24 months, agentic trading will dominate onchain volume.
Not bots. Not scripts.
LLM-driven, context-aware agents making autonomous decisions on risk, capital, and execution.
And they’ll get things wrong.
That’s where we come in.
At @modelguardnet, our first insurance products focus on a single thing:
Coverage for failed autonomous decisions.
• Missed execution
• Protocol misreads
• Logic failures from outdated context
• Adversarial manipulation of agent state
These are novel failure modes. They don’t map to TradFi risk categories — they demand a new underwriting framework built on transparency, verification, and auditability.
ModelGuard is that framework.
We’re building the infrastructure for protocol-native insurance in an agent-first world.
Not off-chain wrappers. Not centralized oracles.
True coverage for agentic risk.
We’ll protect the next generation of capital allocators — even when they don’t have a human in the loop.
@$modelguardnet | $GUARD
1
3
277
Come learn more about ModelGuard tommorow at 10am UTC with @whitneygibbs from @AgentXAVI and @0xcheff_eth from @Wach_AI 🫡
1
3
299
The countdown begins. 6 days until ModelGuard goes live the transparent insurance layer for AI agents. Stay tuned.
3
179
Speaking now with @PlaceholderKey and @xFactorSOL ! Come listen and jam and learn about the upcoming $GUARD
Whassup @virtuals_io fam!
Join @xFactorSOL and I as we talk Virtuals at 11:00am EDT (45 mins).
Also joining us will be the team from @modelguardnet, an insurance layer for the ACP-agentic ecosystem. $GUARD will be launching in 6 days on the Genesis Launchpad.
nitter.app/i/spaces/1DXGyqRjAnPGM
3
248
"Semantic drift" is a form of agentic stress-test; there are millions of different ways to ask an agent for the same thing. We test for overall reliability of outputs.
1
2
272
6/6 The future of agent commerce is insured, transparent, and unstoppable. Ready to build with confidence? Dive in at modelguard.net.
2
179
5/6 For the @virtuals_io community, this is game-changing: standardized trust ratings let virtual worlds, metaverse economies, and DeFi protocols scale without fear. ModelGuard underpins lone avatars and entire agent networks
1
2
191
In ACP, utility trumps hype. The agents that quietly deliver on-chain value are the new MVPs.
1
2
326
Output reliability directly effects the coverage quote we will be offering for different agentic actions. if an agent is found to have a high risk rating, the coverage score will reflect that.
1
2
247
This is part of how we're viewing our initial insurance offerings within Crypto. "not all claims are equal" means we can't evaluate every claim within the same context or information; some will be small, others large. We're starting to figure out exactly what we're going to be able to offer longer-term. Ideally we're able to offer better rates the higher the compliance to cybersecurity standards like AIUC-1, NIST, ISO 42001, etc
1
15