DeFi Users Warned to Revoke Approvals Before Anthropic’s Mythos AI Launches

DeFi Users Warned to Revoke Approvals Before Anthropic’s Mythos AI Launches

Anthropic is planning to make its Mythos AI model available to the public, and a crypto analyst, The DeFi Investor, is advising those involved in decentralized finance to prepare beforehand.

As a crypto investor, I’m a little worried about Mythos. It’s incredibly effective at finding flaws in software, and if it became readily available to everyone, hackers could start finding and exploiting weaknesses in DeFi platforms much faster. That could lead to a lot more attacks and losses for investors like me.

What the DeFi Community Needs to Do

On June 9th, The DeFi Investor shared advice on X (formerly Twitter): to protect their cryptocurrency, users should remove permissions granted to various applications, only use decentralized apps that have been thoroughly checked for security flaws, and divide their funds across multiple wallets to minimize potential losses if one wallet is compromised.

Token approvals are like permissions users grant to apps (smart contracts) to spend their digital tokens. These permissions can build up over time without users realizing it. If one of those approved apps later has a security flaw, it could potentially be exploited, making these approvals a potential security risk.

According to The DeFi Investor, Mythos is remarkably effective at discovering serious security flaws. It’s even claimed to be superior to Claude Opus 4.8, which recently uncovered a critical bug in Zcash.

They added that DeFi will face a huge stress test in the next few months, and indeed, the Zcash vulnerability they mentioned gave a concrete illustration of this.

The privacy-focused cryptocurrency Zcash dropped over 35% in value after an AI-powered security analysis revealed a vulnerability in its system. This flaw could have allowed attackers to create an unlimited number of new ZEC tokens. Following the discovery, prominent crypto investor Arthur Hayes sold all of his Zcash holdings, concerned that the vulnerability might already have been exploited.

Anthropic has limited access to its Mythos AI model to around 50 companies—including tech giants like Amazon, Apple, Google, and Microsoft—since April, as part of a program called Project Glasswing. The goal is to use the model to improve online safety and security. Bloomberg reports that Anthropic now intends to broaden access to 150 additional organizations in 15 different countries.

Several sources, like The Fast Company and reporter Alex Heath, say the publicly released version of Mythos will have significant limitations and won’t be as open as the version available to Project Glasswing collaborators.

A Debate DeFi Was Already Having

The DeFi Investor’s security tips have come at a time when a conversation has been building around the viability of decentralized finance.

In late May, Manuel Aráoz, co-founder of OpenZeppelin, stated that he believes all decentralized finance (DeFi) platforms are currently unsafe. He reportedly recommended that people withdraw their funds from prominent protocols like Aave, MakerDAO, and Compound, because he believes advancements in artificial intelligence have given attackers a significant advantage, making it impossible to guarantee the security of user funds.

Several cryptocurrency projects have been targeted by hackers recently. In April alone, attacks on KelpDAO and Drift Protocol resulted in combined losses exceeding $570 million. Just recently, hackers stole at least $30 million worth of the H token from Humanity Protocol, accessing 17 different wallets.

However, according to Aave Chan Initiative founder Mark Zeller, the fears about AI have been overblown, with fewer than 10% of DeFi security failures in the past year having been caused by code-level vulnerabilities.

According to Bloomberg, Anthropic believes AI will ultimately benefit those who are protecting against its misuse, but the shift will likely be challenging and unstable for a while.

Read More

2026-06-10 00:47