Why We Backed ModelGuard
If 2025 is the year agents get the wheel, ModelGuard is the seatbelt, and every mass-market technology inflection has rewarded the company that sells safety.
The Macro: Autonomy Is Here, but Accountability Isn’t
OpenAI’s product chief recently confirmed that mainstream tools will soon “go off and do a bunch of work” for users rather than wait for prompts. Most of us call it agentic AI. Consumers will simply call it a part of life.
Analysts already predict agents will intermediate everything from shopping trips to cloud infrastructure, potentially reshaping the architecture of the internet itself. These agents will act semi-independently across every domain: financial services, enterprise ops, legal automation, customer support, and everything in between.
But where there’s autonomy, there’s liability. And right now, liability is being punted downstream. Lloyd’s syndicates and Armilla just launched the first insurance policy to cover chatbot errors after a series of embarrassing AI hallucinations. Munich Re’s aiSure program is pitching corporate clients on insurance for model failure. The writing’s on the wall: risk transfer will be a foundational layer of the agentic stack—just as SSL was for e-commerce or audits were for DeFi.
Why ModelGuard: Data, Not Vibes
As agents proliferate, no one knows which ones are safe to trust. The current state of AI deployments is opaque, unaudited, and uninsured. This won’t scale into critical industries.
ModelGuard is building the first fully composable rating and insurance platform for AI agents. It does two things incredibly well:
- It evaluates the quality of an agent before and after launch through real-world stress testing, dynamic probes, and behavioral scoring.
- It underwrites financial risk based on that scoring, offering ACP-compatible insurance pools and real-time alerts when agents drift out of their operating bounds.
From a structural standpoint, it’s Moody’s meets Nexus Mutual. From a market fit standpoint, it’s overdue.
The Genesis Rating System evaluates agents before they launch, scoring team quality, tech feasibility, and token design. The Operational Rating System kicks in after launch, continuously testing agents to verify they actually do what they say they do. Instead of one-time audits, ModelGuard goes further, giving living scores that change as agents evolve.
That data then feeds into ModelGuard’s underwriting engine. Agents with better scores get better rates. Riskier agents pay more—or get denied. Developers can stake $GUARD to back insurance pools, earning yield and participating in upside. And users get alerts when something looks off. It’s a full-stack feedback loop between data, capital, and trust.
The Right Market, the Right Timing
ModelGuard didn’t launch in a vacuum. It embedded itself inside the fastest-growing on-chain AI agent ecosystem: Virtuals Protocol, and is now integrating deeply with its Agent Commerce Protocol (ACP). ACP offers the coordination layer for multi-agent systems, prioritizing utility over hype and enabling autonomous agents to transact, verify, and collaborate seamlessly.
With billions in tokenized agent value and hundreds of new agent proposals, Virtuals ACP is ground zero for composable AI and agent-to-agent communication—and ModelGuard is positioned to become its core security primitive. By aligning with ACP’s standardized smart contracts and built-in evaluation phases, ModelGuard can embed trust directly into the transaction layer.
The team also plans to partner with a growing roster of verification and security agents already building within ACP, extending coverage and strengthening its already-robust security offering.
Today, users chase influencer-built ‘Genesis Tier Lists’ to guide investments. Soon, ModelGuard’s Genesis Ratings will offer a more credible benchmark, and for teams? A prerequisite to launch. The ModelGuard agent is already monitoring top performers in the wild and gathering proprietary risk data before insurance pools have even gone live.
Strategic Alignment with Webspace Labs
At Webspace Labs, we exist to help Web3 AI projects ship fast, launch fairly, and scale safely. We believe the agent economy shouldn’t be gated by closed APIs or opaque standards imposed by megacorps. It should be auditable, decentralized, and permissionless.
ModelGuard builds exactly the kind of primitive we look for: one that introduces trust without requiring trust. By pairing rigorous agent telemetry with protocol-native underwriting, it creates a system where safety becomes composable and capital flows to quality. In other words, ModelGuard does a lot more than just protect the ecosystem. It upgrades it.
Why Now?
The trust gap in AI agents is widening fast. Everyone can deploy an agent. Few can guarantee its performance. Fewer still can price that risk. ModelGuard does all three, backed by data, reinforced by crypto, and governed by a community of risk-aware participants.
This is not a “nice to have.” It’s the missing infrastructure for an entire vertical of technology.
We invested because ModelGuard gives the agent economy its first serious defense primitive, one that makes agents legible, accountable, and insurable. We believe every serious agent protocol will either use ModelGuard or compete with it. And given the pace of adoption and the depth of its data flywheel, we don’t think anyone will catch up.
Website: modelguard.net
X: https://x.com/modelguardnet