# Drift Protocol: why has Claude Mythos made DeFi audits obsolete?

> Is this the end of DeFi audits? Drift Protocol shows how Claude Mythos turns the human factor into the new weak link
> By Covadonga Fernandez · Cybersecurity · 2026-04-17

The $280 million theft from Drift Protocol has been the death certificate of cybersecurity as we know it. While the industry continued to entrust its fate to code audits, the shadow of models like Claude Mythos over this attack, linked to North Korean elite groups, shows that the walls of decentralized finance have an insurmountable crack: the human factor.

When organized cybercrime uses AI to industrialize social engineering, infiltrate critical environments for months and manipulate key actors with a perfection indistinguishable from human, reviewing the smart contract is no longer enough. The Drift case marks a point of no return. The technology that promised efficiency has made a decade's worth of security standards obsolete.

## The dogma of code auditing

For years, the DeFi industry has lived under the dogma of code auditing as a shield against hacks. However, what happened in April 2026 with Drift Protocol has broken that paradigm. With the emergence of Claude Mythos, cybercrime has stopped looking for flaws in Solidity lines to exploit the cracks in human psychology. While auditors review smart contracts, AI is designing social engineering attacks so sophisticated that they make traditional defenses look, overnight, like tools from another era.

The month of April 2026 has become a turning point for the decentralized finance ecosystem. In just a few weeks, a series of coordinated attacks and exploited vulnerabilities have shaken confidence in the sector. At the center of this earthquake is [Drift Protocol](https://x.com/DriftProtocol/status/2040611161121370409), the victim of one of the year's largest thefts, but also a symbol of the entry of artificial intelligence as a cybercrime tool.

## $280 million gone

On April 1, Drift suffered an attack that exposed new forms of vulnerability. Unlike traditional hacks, based on code errors, this was the result of a prolonged social engineering operation. Over months, the attackers managed to infiltrate the protocol's environment, gain the trust of key actors and compromise critical access. The result was devastating: around $280 million disappeared within hours. Several analyses point to the possible involvement of groups linked to North Korea, which reinforces the hypothesis of a highly sophisticated and organized attack.

## [Why Anthropic's AI is today more lethal for the crypto sector than the quantum attack](https://observatorioblockchain.com/en/cybersecurity/la-ia-de-anthropic-es-hoy-mas-letal-para-cripto-que-la-cuantica/)

This case stands out not only for its scale, but for its nature. The failure was not in the code, but in people and processes. In an ecosystem obsessed for years with technical audits, the Drift case shows how threats go far beyond the technical, with the emergence of AI as the protagonist.

## Claude Mythos and audits

Concern in the industry has grown over the potential of advanced models like Claude Mythos and other equivalent systems. These tools can be used to automate social engineering campaigns, generate communications indistinguishable from human ones or even analyze infrastructures in search of weaknesses. In the hands of organized actors, AI not only accelerates attacks, but makes them harder to detect. That is why the combination of AI and cybercrime is beginning to redefine the rules of the game.

The Drift hack is part of a chain of incidents that paint a worrying picture. Earlier in the month, the TMM/USDT liquidity pool on the BNB Smart Chain network suffered a reserve manipulation attack that caused losses of $1.67 million. A few days later, the bridge aggregator Dango lost $410,000 due to a flaw in its smart contract. At the same time, the lending protocol Silo Finance recorded losses due to a misconfiguration of its oracle, while Aethir suffered an access control incident. Even smaller projects, such as Rhea Finance, did not escape this wave, with losses reaching $7.6 million after a coordinated manipulation of its margin system.

## $16 billion stolen in DeFi

Data from the [DefiLlama](https://defillama.com/hacks?time=90d) hack tracker reflect both the historical scale of the problem and its recent acceleration. In total, more than $16 billion has been stolen in the crypto ecosystem over the years, with more than $7.4 billion concentrated in DeFi protocols. But far from stabilizing, the trend is intensifying. In the last 90 days, more than $446 million has been lost, of which more than $424 million corresponds to DeFi. This concentration confirms that decentralized finance remains the main target of attackers.

In this context, North Korea, historically linked to cyberattack operations in the crypto sphere, appears to have evolved its methods. Cases like Drift or the compromise of the Zerion wallet show a trend toward hybrid attacks that combine advanced social engineering with emerging technological tools. AI, in this sense, acts as a capability multiplier, enabling more credible, scalable and harder-to-trace campaigns.

What is happening is not simply a succession of incidents, but a structural transformation of risk in DeFi. For years, the dominant narrative revolved around code security. Today, that narrative is expanding to include human factors, governance and new technologies such as artificial intelligence. Decentralization, one of the sector's pillars, also complicates the response to these attacks, since there is not always an authority capable of intervening quickly.
