Can AI kill DeFi?

Artificial intelligence is rapidly changing cybersecurity - and DeFi is no exception. Recently, Alexandra Gulamova and Igor Gulamov from Savant.Chat, a security firm focused on AI-powered audits, discussed with 1inch how AI is changing the rules for both attackers and defenders. 

AI is making it easier to discover vulnerabilities, combine expertise from multiple domains and uncover attack paths that would have been difficult for even experienced hackers to find just a few years ago. Does that mean decentralized finance is becoming impossible to defend?

Not necessarily. According to Igor and Alexandra, the projects that embrace AI-first security practices will be far better positioned than those relying solely on traditional audits and bug bounties. The question is no longer whether AI will reshape DeFi security - it's whether protocols can adapt quickly enough.

AI is making attackers smarter

One of the biggest misconceptions in the industry is that AI simply makes hacking easier. Igor argues the real change is deeper: AI bridges knowledge gaps that previously limited attackers.

Modern exploits increasingly require expertise across multiple disciplines, from cryptography and mathematics to low-level programming and business logic. Until recently, assembling all that knowledge in a single person - or even a single team - was difficult.

"AI removes these gaps," Igor explains. Instead of needing specialists in every field, attackers can rely on AI to combine expertise across domains, enabling far more sophisticated exploits.

He points to recent zero-knowledge (ZK) exploits as an example. Successfully attacking these systems often requires simultaneously understanding advanced mathematics, cryptography and low-level programming - something AI excels at.

It's not lowering the barrier - it expands what's possible

Alexandra believes that, rather than lowering the barrier to entry, AI dramatically expands what attackers can accomplish.

Previously, hackers needed to assemble people with different specializations. Today, "it could be done by one machine," she says.

The problem is amplified by the fact that many DeFi projects still rely primarily on traditional security approaches - human audits, bug bounty programs and manual reviews. While those practices remain valuable, they are no longer sufficient on their own against AI-assisted adversaries.

Which protocols are most at risk?

According to Igor, the greatest risk lies in systems that combine multiple complex domains.

Protocols involving advanced cryptography, sophisticated business logic, arithmetic-heavy calculations or cross-domain interactions become significantly more attractive targets because AI can reason across all these areas simultaneously.

Bridge protocols remain especially exposed as well. Recent attacks have demonstrated how attackers can combine seemingly unrelated weaknesses - such as arithmetic bugs with flaws in business logic - to create devastating exploits. AI makes identifying these combinations considerably easier.

He also notes that closed-source software is no longer much of a defense.

"Closed source isn't a shield," he says, because AI-powered reverse engineering and decompilation continue to improve.

Can AI defend DeFi too?

Fortunately, the same technology empowering attackers can also strengthen defenders.

Igor believes AI's greatest value isn't monitoring live smart contracts in real time, where defenders inherently face an asymmetric challenge. Attackers can spend weeks using enormous computing resources to discover an exploit before launching it in seconds.

Instead, AI delivers its biggest advantage much earlier - in development.

Running comprehensive AI audits throughout continuous integration and continuous deployment (CI/CD), scanning every pull request and identifying broken invariants before code reaches production dramatically reduces the available attack surface.

As Igor explains, defenders don't necessarily need to discover the exact exploit an attacker would eventually use. If AI identifies a broken invariant, developers can simply fix it before anyone has a chance to weaponize it.

Human auditors aren't going away

Neither expert believes AI will replace human security experts anytime soon.

Alexandra describes AI audits as a new foundational layer rather than a replacement for traditional reviews.

Her recommended workflow is straightforward:

• Run comprehensive AI audits across the codebase.
• Integrate AI into every CI/CD pipeline and pull request.
• Periodically re-scan the entire codebase as it evolves.
• Finish with human security audits before major releases.

Each layer catches different classes of issues, making the overall security posture significantly stronger.

The race between sword and shield

Looking ahead, Igor expects an ongoing arms race.

Attackers will inevitably gain access to increasingly powerful models - including open-source models with fewer restrictions. As frontier AI becomes widely available, sophisticated vulnerability discovery will become accessible to many more threat actors.

That means defenders cannot afford to rely on outdated tooling or AI models that lag months behind the state of the art.

Instead, security teams will increasingly combine frontier LLMs with formal verification, invariant checking and continuous AI-driven auditing to eliminate vulnerabilities before attackers find them.

So... can AI kill DeFi?

Both experts answer with a simple and confident "No."

AI undoubtedly gives attackers new capabilities. It can combine knowledge across disciplines, uncover complex exploit chains and dramatically accelerate vulnerability discovery.

But AI also gives defenders unprecedented tools to secure protocols before they're deployed.

The winners won't be the projects trying to fight AI - they'll be the ones embracing it first. As AI reshapes both sides of cybersecurity, DeFi's future won't depend on whether artificial intelligence exists. It will depend on how effectively protocols use it to stay one step ahead of attackers.

For more security insights subscribe to our newsletter

Disclaimer: This content is for general information purposes only and does not constitute financial, investment, tax or legal advice and is not a recommendation to buy or sell any particular digital asset or to employ any specific investment strategy.