Crypto: Avalanche Founder Warns of Vulnerabilities That AI Could Exploit

The founder of Avalanche believes that the crypto industry is too focused on a potential breakdown of cryptography. According to him, artificial intelligence could first exploit the existing flaws in software and wallets.

In Brief

• Emin Gün Sirer considers the debate on AI and crypto security largely exaggerated and points first to software bugs.
• He believes that the code of Ripple and other networks could hide flaws that AI will detect more easily than humans.
• No specific vulnerabilities have been reported in this debate, which was initiated on October 7 by researcher Justin Drake.

The Threat of AI Reignites the Debate on Ethereum Security

The debate gained momentum on October 7, when Justin Drake, a researcher at the Ethereum Foundation, urged the industry to prepare a "bunker mode." He specifically recommends transferring funds to addresses that have never been used, in order to keep public keys hidden behind a hashing function.

His concern revolves around ECDSA, the signature mechanism used by many wallets. According to his scenario, mathematical advances accelerated by AI could weaken this system in a few months, rather than over several years.

This hypothesis has sparked mixed reactions. Yehuda Lindell from Coinbase has strongly criticized this perspective, while the technical director of Ledger has highlighted the risks associated with a hasty migration. Vitalik Buterin acknowledges the existence of a serious risk without calling for urgent action.

Emin Gün Sirer Focuses First on Software Flaws

On October 9, the founder of Avalanche took a stand on X. "The debate about the threat posed by AI to crypto is mostly just talk," he writes. In his view, AI will find flaws in existing systems long before breaking ECDSA. These vulnerabilities could allow attackers to steal billions of dollars from software libraries and wallets.

Sirer specifically cites Ripple and other networks that concentrate significant value without undergoing thorough code review. According to him, believing that no flaw difficult for a human to spot escapes developers is an illusion. AI could indeed facilitate the discovery of these vulnerabilities, sometimes unknown to publishers and exploited before their correction.

However, Sirer does not completely dismiss the cryptographic risk. He deems it relevant to prepare backup solutions and indicates that Avalanche is exploring hashing-based alternatives. Nevertheless, he believes that securing existing code should come first.

Recent incidents reinforce his argument. The crypto sector lost $766 million in September. Additionally, a security team demonstrated this summer that an AI could detect 85 critical flaws.

Cryptography or Bugs: Two Risks to Anticipate

These two threats remain distinct. A breakdown of ECDSA could potentially compromise the security of many cryptographic keys. In contrast, software flaws already expose users to attacks, without waiting for a mathematical breakthrough.

The issue also concerns Bitcoin: over six million bitcoins would rest on public keys already visible, representing more than 31% of the circulating supply. The migration recommended by Drake aims to reduce this exposure, but it does not resolve the vulnerabilities of the software.

The debate remains open. While waiting for potential advances capable of undermining cryptographic signatures, developers must also strengthen the systems that protect funds.