Rapid progress in artificial intelligence models may have tipped the cybersecurity balance in favor of attackers. Mitchell Amador, Chief Executive of bug bounty platform Immunefi, said in his evaluation at the WAIB Summit in Monaco that new generation artificial intelligence has increased attacks in the field of decentralized finance again.
Increased risk of attack with artificial intelligence
According to Amador, the proliferation of new models, especially Claude Opus 4.8 and ChatGPT 5.5, was effective in the re-acceleration of attacks against crypto asset platforms in 2026. According to him, this process paved the way for software vulnerabilities to be found faster and for malicious actors to act more effectively.
Immunefi CEO Mitchell Amador stated that there is a critical period ahead for the crypto industry, and security pressure may continue until defense teams use the same artificial intelligence tools to their advantage.
Industry-wide attacks rose sharply in April 2026, according to DefiLlama data. Over $634 million in assets were stolen from crypto platforms during this period. This figure was recorded as the highest monthly level since the Bybit attack in February 2025, bringing total losses to approximately 1.4 billion dollars.
Mini dictionary: DefiLlama is an on-chain data platform that compiles total locked assets, transaction volume and attack data on decentralized finance protocols. It is among the frequently used sources in the industry, especially for comparative data regarding the DeFi ecosystem.
The coming years are seen as critical
Amador stated that the next three to four years will be a kind of endurance test for the crypto industry. He said that during this period, defense teams should develop code infrastructures that are strong enough for attackers to overcome, using the protective capabilities of the same artificial intelligence tools.
He stated that this period could be reduced to less than two years if more community-contributed security solutions are adopted. According to the assessment, the current pressure may weaken as security researchers begin to use artificial intelligence more effectively on the defense side.
DeFi attacks are on the agenda again
Amador’s statements came right after Anthropic announced the new Claude Mythos model called Fable 5. Concerns have been expressed in the industry about the possibility of this model accelerating crypto exploits. Anthropic, on the other hand, announced that topics such as cyber security were directed to a different model, Claude Opus 4.8, and that there are various protections in this area.
Security discussions have become even more visible due to the recent major DeFi vulnerabilities. On April 19, an attacker withdrew approximately 116,500 staked Ether from the rsETH bridge using the Kelp DAO’s LayerZero infrastructure. At that time, the value of these assets was estimated to be between $290 million and $293 million.
Mini dictionary: DVN, or decentralized validator network, is a security layer used to verify cross-chain messages. In this structure, if verification remains too dependent on a single path or a single verifier, a single point of failure may occur in the system.
LayerZero noted that Kelp DAO’s 1/1 DVN configuration creates a single point of failure by relying on a single validator path for cross-chain messages. The company stated that this installation was not recommended before. The incident brought the relationship between architectural choices in protocols and security consequences to the agenda again.
