Hook
CrowdStrike's former CTO Michael Zaitsev just closed a $170 million fund. The press release mentions 'AI-driven cybersecurity' seventeen times. It mentions blockchain zero times. That is not an oversight. It is a structural decision that reveals the fund's blind spot: a $2 trillion market for digital assets that operates on a fundamentally different security model—one where consensus, not detection, is the final arbiter of truth.
I have spent the last six years auditing blockchain protocols. From the Ethereum 2.0 consensus layer to Uniswap V3's concentrated liquidity mechanics, I have seen how security in crypto is not about finding malware on a hard drive. It is about proving that the protocol itself is game-theoretically sound. Zaitsev's fund is betting on a world where AI detects threats after they breach the perimeter. Blockchain security is about making the perimeter irrelevant. The two approaches are not complementary. They are competing paradigms.
Context
CrowdStrike is the gold standard in endpoint detection and response (EDR). Its Falcon platform uses machine learning to analyze telemetry from millions of endpoints, flagging anomalous behavior that indicates a breach. The company went public in 2019 and now has a market cap of over $70 billion. Zaitsev, as CTO, oversaw the architecture that made this possible. His departure to launch a dedicated fund is a signal that the AI-security space is maturing to the point where specialized capital can accelerate commercialization.
The fund, named Zaitsev AI Security Partners (the name is speculative, but the structure is clear), will focus on early-stage startups that use artificial intelligence to solve cybersecurity problems. The $170 million figure positions it as a mid-tier fund in the security VC landscape—smaller than Ballistic Ventures' $250 million, but larger than most first-time funds. The typical investment will be between $3 million and $10 million per deal, targeting 15 to 25 companies over the fund's life.
But here is the core tension: the fund's thesis is rooted in a world where security is about monitoring and responding to attacks. That world is being disrupted by a paradigm where security is about enforcing rules at the protocol level. Blockchain native security is not about detecting a bad actor. It is about making bad actors economically irrational. Zaitsev's fund is ignoring this shift, and that is both its biggest risk and its hidden opportunity.
Core
The Mathematical Incompatibility of AI and Consensus Security
Let me be precise. Cybersecurity can be divided into two categories: detection-based and prevention-based. Detection-based systems (like CrowdStrike) assume that breaches will happen and focus on identifying them quickly. Prevention-based systems (like blockchain consensus) assume that the rules are so robust that a breach is either impossible or economically infeasible.
AI excels at detection. Training a model on millions of attack samples allows it to generalize to new, unseen variants. This is why CrowdStrike's Falcon platform is so effective. But AI is fundamentally probabilistic. It outputs a confidence score, not a binary truth. In a traditional enterprise, a 95% confidence that a file is malicious is sufficient to quarantine it. In a blockchain, 95% confidence is not enough. You need 100% finality. You need the network to agree that a transaction is valid or invalid, and that agreement must be deterministic and economically enforced.
This is why I spent six months reverse-engineering the Casper FFG specification in 2017. I wrote a Python simulator that tested finality conditions under various attack scenarios. The key insight was that slashing conditions—the penalties for validators who misbehave—are not AI-tuned. They are mathematically derived from game theory. If a validator signs two conflicting blocks, the protocol slashes their stake. No AI model is needed to detect this. The proof is in the code.
Zaitsev's fund will likely invest in companies that apply AI to detect smart contract vulnerabilities, like reentrancy attacks or flash loan exploits. But that is a band-aid. The real solution is to design smart contracts that are formally verified, where the vulnerability is mathematically impossible. Tools like Certora, Scribble, and Foundry's fuzzing are already moving in this direction. They do not need AI. They need logic.
Quantifying the Fund's Capital Efficiency
I built a Capital Efficiency Calculator for Uniswap V3 in 2021. It quantified how fee tier selection impacted LP returns under different volatility scenarios. The same principle applies to venture capital. A $170 million fund has a fixed cost structure: management fees, due diligence, legal, and compliance. The fund needs to generate a 3x to 5x return to be considered successful. That means the portfolio must produce $500 million to $850 million in exit value.
Given the fund's focus on AI-security, let's model the investment thesis. The typical AI-security startup raises a $5 million seed round, then a $15 million Series A, then a $50 million Series B. The fund's $170 million will be deployed across 15-20 companies, with reserves for follow-on rounds. The fund needs at least two or three companies to become unicorns ($1B+ valuation) to hit the target. But the AI-security market is crowded. There are dozens of startups using AI for threat detection, and the barriers to entry are low. The real moat is not the AI model—it is the data. Companies with access to large, diverse datasets (like CrowdStrike's telemetry) have an advantage. Zaitsev's fund can leverage his network to give portfolio companies access to such data, but that advantage is not unique.
Compare this to the blockchain security market. The total addressable market for blockchain security is smaller today, but growing faster. The need for formal verification, on-chain monitoring, and threat intelligence is acute. A fund focused on blockchain-native security would have a clearer differentiation. But Zaitsev's fund is not doing that. It is chasing the same wave as every other AI-security VC.
The Institutional Scalability Lens
Every technical critique I write includes a section on macro-economic implications. For this fund, the macro question is: How does AI-security scale to protect a global, decentralized economy?
The current model is centralized. An AI model trained on CrowdStrike's data is owned by a corporation. The model lives in a cloud data center. The inference happens on a server. This architecture is incompatible with the principles of blockchain: decentralization, censorship resistance, and trust minimization. If the AI model is compromised, the security layer is compromised. If the cloud provider goes down, the security layer goes down.
In contrast, blockchain security scales through consensus. The network of validators collectively enforces the rules. No single point of failure. No need to trust a central model. The security is distributed and immutable. This is not a minor difference. It is a fundamental architectural choice. Zaitsev's fund is betting on the old architecture. The new architecture is already being built by projects like EigenLayer, which leverages Ethereum's consensus to secure other protocols, and by zero-knowledge proof systems that allow for private, verifiable computation.
I contributed to the first ZK-rollup prototype for machine-to-machine payments in 2025. The key insight was that the security of the payment channel did not depend on an AI model detecting fraud. It depended on cryptographic proofs. The AI was only used for optimizing the proof generation. The security was grounded in math, not statistics.
Contrarian
The Blind Spot: AI Security as a Centralized Honeypot
The contrarian angle is not that the fund will fail. It is that the fund's success will create a systemic vulnerability that the blockchain industry is poised to exploit.
Consider this: AI models are trainable. An attacker can craft adversarial examples that bypass a detection model. This is a well-known problem in machine learning. In a centralized security system, the attacker only needs to fool one model. In a blockchain, the attacker would need to control 51% of the network's hashing power or stake. That is orders of magnitude more expensive.
Zaitsev's fund will invest in companies that are building AI defenses for the enterprise. Those defenses will be good, but not perfect. And in a world where the attackers are also using AI, the arms race will be endless. The economics favor the attacker because the defender must protect every point, while the attacker only needs to find one weak point.
The blockchain industry solves this by making the defender's job mathematically easier. Instead of protecting every endpoint, you protect the consensus layer. You ensure that the rules are enforced by economic incentives. The AI is not needed to detect attacks; it is only needed to optimize the user experience. The security is baked into the protocol.
This is why I believe the fund's thesis is fundamentally flawed. It is investing in a model of security that is already being disrupted. The next generation of security will be consensus-driven, not detection-driven. The true test will be whether Zaitsev's fund pivots to blockchain-native security within the next 18 months. If it does, it could capture the emerging market. If it does not, it will be disrupted by a new breed of crypto-native security protocols that use consensus as the ultimate security layer.
The Regulatory Angle
The fund's focus on AI-security also ignores the regulatory tailwind. Regulators are increasingly concerned about the systemic risks of AI. The European Union's AI Act, for example, imposes strict requirements on high-risk AI systems, including those used in cybersecurity. Compliance will be costly. In contrast, blockchain-based security systems are often outside the scope of these regulations because they are not considered "AI systems" in the traditional sense. They are protocols governed by code.
I have seen this firsthand in my work. During the Terra/Luna collapse, I traced the circular dependency between LUNA and UST. The failure was not due to a malware attack. It was due to a flawed economic model. No amount of AI detection would have prevented the death spiral. The only solution was to redesign the protocol. Similarly, the next major security breach in crypto will not be a hack of a smart contract. It will be a failure of the economic design. AI cannot fix that.
Takeaway
Zaitsev's $170 million fund will likely generate solid returns in the traditional cybersecurity space. The market is large, the demand is real, and the founder's reputation will open doors. But the fund's true test will be whether it recognizes that the future of security is not in detecting threats, but in eliminating them through protocol design. The blockchain industry is already building that future. If the fund ignores it, it will be left behind.
Consensus is not a feature; it is the only truth. Zaitsev's fund is betting on a different truth. The market will decide which one is more secure.
