AI and Cybersecurity: An Ongoing Battle in Decentralized Finance

Recent advancements in artificial intelligence (AI) appear to have tilted the cybersecurity scales in favor of cybercriminals, as discussed at the WAIB Summit in Monaco. The conversation, led by Mitchell Amador, CEO of Immunefi, zeroed in on the reinvigorated pattern of attacks in decentralized finance (DeFi) sectors caused by the current generation of AI models.

The increasing availability of sophisticated AI frameworks, such as Claude Opus 4.8 and ChatGPT 5.5, has enabled attackers to discover and exploit vulnerabilities at a much faster rate. As Amador highlighted, these tools provide a strategic advantage to cybercriminals, who can now reveal and target protocol flaws with heightened efficiency.

DefiLlama reports that the industry saw a sharp increase in hacks in April 2026, leading to losses that exceeded 634 million dollars. This figure not only underscores a growing security problem but also echoes the notorious Bybit hack of February 2025, which culminated in damages approaching 1.4 billion dollars.

How Can AI Improve Security?

Amador views the coming years as pivotal for the crypto industry’s security landscape. He suggests that only by embracing AI defenses can developers remain a step ahead of cyber adversaries. The task is to create secure codebases that leave potential attackers at bay.

Community-driven solutions could significantly shorten this arduous phase. Amador proposes that with more widespread adoption of such initiatives, the looming threat environment might begin to dwindle as researchers become adept at utilizing AI for protection.

This dialogue occurred alongside Anthropic’s launch of its new AI model, Claude Mythos (Fable 5), further escalating concerns over the frequency of crypto-related breaches. As Anthropic noted, Claude Opus 4.8, equipped with specific safeguards, handles security-focused AI tasks separately.

In April, a security lapse involving Kelp DAO highlighted the ongoing peril for DeFi platforms. Exploiting LayerZero’s infrastructure, an attacker siphoned off approximately 116,500 restaked Ether, valued between 290 and 293 million dollars. Such incidents stress the urgent need for re-evaluating security protocols and design choices.

– The introduction of new AI models by companies like Anthropic adds complexity to existing cybersecurity dynamics.
– Kelp DAO’s incident underscores the risks associated with reliance on single-path validator networks.
– Calls for community-driven security increase as AI continues to evolve both threat and defense landscapes.

LayerZero pointed out that the Kelp DAO breach resulted from its 1/1 DVN setup, which overly depended on a single validator. They had previously advised against this configuration, prompting intense discussions on the intricate link between protocol architecture and security decisions.