The $27 Million Bet on a Security Gap That Shouldn't Exist

Blackstone just committed $27 million to a company called Huskeys. The premise: AI agents need their own security layer. The math is perfect; the reality is broken. Let me be precise about what this means. Huskeys is building technology to monitor and secure network traffic generated by AI agents. Not human traffic. Not traditional API calls. Agent-to-agent communication. Machine-to-machine chatter. The kind of traffic that doesn't exist at scale today but will explode as autonomous systems proliferate. Gartner projects that by 2028, at least 15% of daily work decisions will be made autonomously by AI agents. In 2024, that figure was under 5%. The trajectory is clear. The security implications are not. Traditional security infrastructure was designed for human behavior patterns. Firewalls, intrusion detection systems, signature-based threat hunting—all built around how people interact with networks. AI agents don't behave like people. They execute high-frequency, automated, multi-step operations. They call APIs dynamically. They make decisions based on context. They don't follow predictable patterns. Huskeys is betting that this gap between traditional security and agent-native traffic creates a new market. Blackstone is betting that Huskeys can capture it. Here's what bothers me about this deal. The technology is unproven. The company has disclosed no technical architecture, no detection algorithms, no enterprise customers. The $27 million A round suggests they've moved past seed validation, but they're still in the POC-to-production transition. The technical approach hasn't converged. There are multiple competing paradigms—traffic analysis, behavior modeling, LLM-assisted detection—and no clear winner has emerged. I've seen this pattern before. In 2021, I audited a smart contract project that raised $30 million on the strength of its marketing narrative. The code had a critical integer overflow vulnerability in the staking reward calculation. I flagged it. The team dismissed it as a theoretical edge case. The exploit was triggered within 48 hours of launch. $28 million drained. Code is the only honest actor. Everything else is narrative. Let me quantify the market opportunity, because that's where the real analysis lives. The global AI software market is projected to exceed $300 billion by 2026. Security spending typically represents 3-5% of IT budgets. If AI agent security captures even a fraction of that, we're looking at a $5-15 billion addressable market by 2028. That's enough to support multiple unicorns. The customer profile is compelling. Financial institutions, healthcare providers, government agencies—these are the organizations deploying AI agents most aggressively. They also face the strictest compliance requirements. They have high willingness to pay and high security standards. The procurement cycles are long, but the contracts are sticky. Security products typically see renewal rates above 90%. But here's the hidden cost that nobody in the press release is talking about. The cloud providers are watching. AWS, Azure, and Google Cloud are all building AI agent platforms. Amazon Bedrock Agents. Azure AI Foundry. These platforms will inevitably embed security capabilities as native features. If they do, independent security vendors like Huskeys face an existential threat. This is the classic innovator's dilemma inverted. The incumbents—Palo Alto Networks, CrowdStrike, Zscaler—are adding AI capabilities to their existing products. They're treating AI as an enhancement to traditional security, not a fundamentally new paradigm. That gives Huskeys a window. But the cloud providers are different. They control the infrastructure layer. They can bundle security into the platform at near-zero marginal cost. I've seen this movie before. In 2023, I analyzed the gas fee structures on Uniswap v3. I bypassed the standard UI and interacted directly with the mempool. What I found was that 40% of transaction costs weren't fees at all—they were MEV bribes paid to validators. For every $100 a user paid, only $3 went to liquidity providers. The rest was siphoned by bots. The protocol was extractive, not additive. The industry narrative was designed to obscure that reality. Huskeys faces a similar structural challenge. The security layer they're building is valuable only if it remains independent. The moment a cloud provider decides to bundle agent security into their platform, the standalone market collapses. The question isn't whether that happens. It's when. Let me look at the valuation math. A $27 million A round with Blackstone as the lead investor. Assuming they took 15-25% equity, that puts Huskeys at a post-money valuation between $108 million and $180 million. For comparison, Protect AI raised a B round at roughly $400 million valuation. TrojAI raised an A round at around $50 million. Huskeys sits in a reasonable range for the sector. The burn rate is the real constraint. A typical Israeli security startup with 30-50 employees runs $8-12 million annually. That gives Huskeys a 2-3 year runway. Enough time to iterate on the product and land initial customers. Not enough time to survive a prolonged enterprise sales cycle if the product doesn't resonate immediately. Here's what the bulls get right. The timing is actually good. AI agents are moving from technical validation to production deployment. Enterprises are starting to ask hard questions about security. The market is about to experience a demand spike, and there's no established player with a dedicated solution. Blackstone's involvement matters beyond the capital. They manage over $1 trillion in assets. Their portfolio includes dozens of enterprise software companies that are potential Huskeys customers. The strategic value of that ecosystem access far exceeds the $27 million check. And the Israeli security ecosystem is a genuine advantage. The military intelligence background—Unit 8200 and its alumni—creates a talent pool that's uniquely suited to this kind of work. Check Point, CyberArk, Wiz—they all came from the same environment. The technical DNA is real. But I keep coming back to the fundamental uncertainty. The technology is unproven. The competitive landscape is fluid. The cloud provider threat is existential. And the company has disclosed almost nothing about how their solution actually works. Trust is a variable that must be zero. Every transaction is a potential extraction point. Let me be clear about what this deal actually signals. It's not a bet on Huskeys specifically. It's a bet on the category. Blackstone is saying that AI agent security is a real market that will emerge regardless of which company wins. They're buying exposure to the theme at a reasonable price. That's a defensible strategy. But it doesn't tell you anything about whether Huskeys will be the winner. The signals I'm watching are concrete. Does Huskeys publish a technical whitepaper in the next six months? Do they announce enterprise customers with real deployment case studies? Do the cloud providers release native AI agent security features? Each of these data points will tell us more than the $27 million ever could. The illusion breaks when the liquidity dries up. The question is whether Huskeys can build a moat before that happens. Logic holds; incentives collapse. The incentives here are aligned for the category, not necessarily for the company. Blackstone gets portfolio diversification. Huskeys gets capital and credibility. The market gets a signal that AI agent security is investable. What we don't get is evidence that the technology works. And in security, that's the only thing that matters. Between the commit and the block lies the trap. Between the press release and the production deployment lies the truth. I'll wait for the data.