Artificial intelligence is changing the economics of blockchain security, giving attackers the ability to find software weaknesses faster than many traditional review processes can catch them, NEAR Protocol co-founder Illia Polosukhin said in recent remarks.
Polosukhin said the growing use of AI-assisted hacking means blockchain developers may need to rely more heavily on formal verification, a mathematical method used to prove that software behaves as intended. The approach, once viewed as too costly and slow for many crypto projects, is becoming more practical as AI systems begin handling large parts of the proof-writing process.
His comments come as blockchain networks, decentralized finance applications and digital asset platforms face increasing pressure to strengthen their security models. The industry has suffered repeated exploits in recent years, with attackers using more advanced tools, faster automation and deeper knowledge of smart contract design. Polosukhin argued that manual audits and expert code reviews, while still useful, are no longer enough on their own.
He pointed to the Balancer incident in October of last year as one of the first visible examples of AI being used to help identify and exploit code vulnerabilities. According to Polosukhin, that episode showed how machine-learning systems can speed up the process of discovering flaws in complex software.
The warning is especially significant because blockchains often hold large amounts of value in open-source code. Unlike traditional financial systems, where internal controls and legal recovery processes can sometimes reduce damage after a breach, many blockchain transactions are irreversible. Once funds are drained from a vulnerable contract, recovering them can be difficult or impossible unless the attacker returns them, law enforcement intervenes successfully, or a protocol has built-in emergency controls.
Polosukhin said the solution is not to trust AI blindly, but to use it to automate parts of a process that can still be independently checked. Formal verification produces mathematical proofs that can be reviewed and validated, making the method different from relying on an AI model’s judgment alone.
“AI can generate the proof, but the proof itself can be verified,” he said in substance, describing the technology as a tool for accelerating security work rather than replacing trust with another opaque system.
Why formal verification is gaining attention
Formal verification is a long-established field in computer science. It uses mathematical reasoning to prove that software satisfies specific rules or properties. In blockchain, those properties might include proving that tokens cannot be created out of thin air, that a transaction system balances correctly, or that a protocol cannot enter an unsafe state under certain conditions.
For years, formal verification was considered valuable but difficult to apply broadly. It required specialized knowledge, significant time and careful modeling of a system’s expected behavior. In fast-moving crypto markets, many projects favored manual audits, bug bounties and rapid deployment cycles instead.
Polosukhin said that balance is changing. He argued that modern AI models can now draft much of the proof work that previously required large teams and long timelines. Processes that once took several years may now be shortened to weeks in some cases, depending on the size and complexity of the codebase.
That does not mean AI can automatically secure a project. A formal proof is only as good as the property it is proving. If developers define the wrong assumptions, omit important risk conditions, or fail to include real-world system behavior, a mathematically verified component may still leave gaps elsewhere. However, when used correctly, formal verification can reduce entire classes of errors that are difficult to catch through manual review alone.
The appeal is clear in blockchain because smart contracts are public, composable and often targeted immediately after launch. Attackers can inspect code, simulate transactions and search for weaknesses without needing physical access to a system. AI tools can make that process faster by assisting with pattern detection, exploit generation and code analysis.
This creates a race between offense and defense. If attackers use AI to scale vulnerability discovery, defenders may need comparable automation to keep up.
The Zcash Tachyon example
Polosukhin cited the Zcash-related Tachyon project as a recent example of AI-assisted formal verification in practice. The Tachyon team used AI to help produce a mathematical validation for Zcash’s Ironwood system, confirming that it avoids an undetected counterfeiting issue previously identified in the Orchard network.
The significance of the example lies in the type of risk being addressed. In privacy-focused digital currency systems, counterfeiting bugs can be especially dangerous because transaction details are intentionally shielded. If a system has a flaw that allows coins to be created without detection, the damage may not be immediately visible to users, developers or market participants.
Tachyon’s work showed how AI-generated proofs can speed up a deeply technical verification process while still leaving room for independent checking. The team reported that AI-driven verification compressed work that previously could have taken years of manual effort.
For developers, that matters because the security burden in blockchain is rising. Protocols are expected to add new features, handle more activity, interoperate with other chains, and maintain strong protections against economic and technical attacks. Each added feature can create new complexity. Formal verification offers a way to reason about that complexity before an exploit occurs.
Polosukhin suggested that the same model should not be limited to blockchains. He said broader internet software and infrastructure could also benefit from AI-assisted mathematical verification. That includes systems where failure can have major consequences, such as authentication tools, cloud infrastructure, payment systems, embedded devices and critical communications software.
The broader argument is that modern software systems have become too complex for human review alone. AI may be part of the reason attackers are getting faster, but it may also be part of the answer for defenders.
NEAR adds security-focused upgrades
On the same day as Polosukhin’s remarks, NEAR announced a platform upgrade that includes support for quantum-resistant cryptographic keys and automatic capacity scaling. The changes are intended to prepare the network for future risks tied to both cybersecurity and computing power.
Quantum-resistant keys are designed to protect against potential future machines that could break widely used cryptographic systems. Large-scale quantum computers capable of defeating current cryptography do not yet exist in a commercially practical form, but many networks and technology providers are preparing early because cryptographic transitions can take years.
For blockchains, the concern is straightforward. A public blockchain depends on cryptographic signatures to prove control of accounts and authorize transactions. If older signature schemes become vulnerable in the future, users and protocols would need secure migration paths. Adding support for quantum-resistant keys is one way to begin that transition before the threat becomes immediate.
The second part of the NEAR upgrade, automatic capacity scaling, is aimed at network performance. As activity rises, blockchain networks can face congestion, slower confirmations and higher transaction costs. Automatic scaling is intended to help the protocol adjust capacity when demand increases.
For traders and application users, scaling can affect the practical experience of using a network. During periods of intense market activity, on-chain systems can become more expensive or less predictable. A network that adjusts capacity more smoothly may be better positioned to maintain usability during spikes in demand.
Still, performance improvements and new key formats do not remove the need for careful security design. Protocol-level upgrades can reduce certain risks, but application code, bridges, wallets and smart contracts remain major attack surfaces.
AI changes the speed of attacks
The central concern raised by Polosukhin is speed. Traditional blockchain security has relied heavily on human experts reviewing code before deployment. These reviews often include manual inspection, automated testing, economic modeling and bug bounty programs. The process can identify serious flaws, but it has limits.
AI systems can examine large codebases quickly, compare patterns across known vulnerabilities and help generate test cases. In the hands of defenders, this can improve security coverage. In the hands of attackers, it can reduce the time needed to move from code review to exploit.
That shift is important because many crypto projects operate in public. Open-source development is a strength for transparency, but it also gives attackers access to the same code that auditors and community researchers examine. If a vulnerability exists, the key question becomes who finds it first.
Polosukhin’s view is that formal verification may become a necessary baseline for high-value systems. The more value a protocol holds, the more attractive it becomes as a target. If AI lowers the cost of finding vulnerabilities, attackers can scan more protocols with less effort.
This does not mean every project can formally verify every line of code immediately. But critical components, such as token minting logic, bridge contracts, consensus mechanisms and privacy systems, are likely to face growing expectations for stronger mathematical guarantees.
Security failures remain costly
The digital asset sector has already seen the consequences of weak security. Billions of dollars have been stolen from crypto-related platforms, bridges, wallets and decentralized finance applications in recent years. Some attacks have exploited smart contract bugs. Others have involved compromised private keys, social engineering, flawed access controls or weaknesses in infrastructure rather than blockchain code itself.
One of the largest single incidents involving a trading platform cost users and market participants roughly $1.5 billion, underscoring how quickly losses can escalate when attackers breach a high-value target. Across the wider digital asset space, annual theft totals have reached into the billions, reflecting both the financial incentive for attackers and the difficulty of securing open, programmable financial systems.
These figures have shaped how traders evaluate protocols. Security reports, audit histories, bug bounty programs, upgrade controls and proof systems are becoming part of the due diligence process. A project’s market narrative may draw attention, but its security model can determine whether it survives a serious attack.
Formal verification is not a guarantee against all losses. It cannot prevent every phishing attack, private-key theft, governance failure or malicious insider event. It also cannot prove that a system is safe unless the right properties are modeled. But it can make certain types of software failure much harder to miss.
A shift from trust to proof
Polosukhin’s argument reflects a broader change in blockchain security culture. The industry has long used the phrase “don’t trust, verify” to describe the idea that users should be able to inspect and validate systems rather than rely on institutions. Formal verification extends that idea from transaction history to software behavior.
Instead of asking users to trust that a smart contract works because a team says it does, mathematical proofs can show that specific parts of the system satisfy defined rules. That distinction matters as protocols become more complex and interconnected.
AI complicates the picture. On one hand, it introduces new threats by helping attackers automate discovery. On the other, it can make high-end verification methods more accessible. The outcome may depend on how quickly developers adopt defensive uses of the same technology.
For blockchain teams, the message is becoming clearer: security practices that were considered advanced several years ago may soon become standard expectations. Manual audits, while important, may need to sit alongside machine-assisted testing, continuous monitoring, formal proofs and stronger cryptographic design.
For traders, the practical takeaway is not to assume that a protocol is safe simply because it is popular, fast or widely discussed. Public audits, verification reports, key-management practices and upgrade mechanisms can provide useful signals about how seriously a project treats security.
Polosukhin’s remarks also point to a future where software security is less about one-time reviews and more about continuous proof. As code changes, proofs may need to be updated. As new threats emerge, assumptions may need to be tested again. As AI systems improve, both attackers and defenders will likely become faster.
The blockchain sector has often moved quickly, sometimes at the expense of operational caution. The rise of AI-assisted exploit discovery may force a more disciplined approach. If attackers can find weaknesses in minutes or hours, security teams cannot rely on processes that take months or years without automation.
NEAR’s latest upgrade shows one response to that pressure: prepare for stronger cryptography, build for heavier usage and treat future computing threats as current design problems. Polosukhin’s broader point is that the same thinking should apply to software correctness itself.
In a market where code can control billions of dollars, trust is becoming a weaker defense. Proof, especially when accelerated by AI and checked independently, may become one of the industry’s most important security tools.
Strengthen your security-first strategy with AI—discover how AI complements blockchain to protect protocols from evolving exploit techniques.
Disclaimer: The content on this page is provided for general informational purposes only and does not represent the views or financial advice of Toobit. We make no guarantees regarding the accuracy or completeness of this information and shall not be held liable for any errors, omissions, or outcomes resulting from its use. Investing in digital assets involves risk; users should independently evaluate their financial situation and the risks involved. For further details, please consult our Terms of Service and Risk Disclosure.
