Ethereum co-founder Vitalik Buterin publicly disagreed with an investor's prediction that artificial intelligence would crash Bitcoin's price by more than 50% within two years, setting off an exchange on X that drew enough attention to shift the forecaster's stated confidence level within hours.
Liron Shapira posted on September 7 that he assigned 50% confidence to Bitcoin falling more than 50% over the next two years because of AI undermining what people believe are its security or robustness guarantees. Buterin responded the same day, taking the opposite position and explaining his reasoning across a single post.
"My basic reasons are that I am quite optimistic about cybersecurity in the long term and I see the primary problem as being getting the transition," Buterin wrote, "and I expect BTC to handle at least any issues that do not require social consensus well (upgrading clients, mining pools, etc to deal with network-layer hacks is in this category) (and I think the probability of actual breaks on hashes or PoW is tiny)."
I take the opposite side of that.
— vitalik.eth (@VitalikButerin) September 7, 2026
My basic reasons are that I am quite optimistic about cybersecurity in the long term and I see the primary problem as being getting the transition, and I expect BTC to handle at least any issues that do not require social consensus well…
Buterin added that he was already effectively taking the bet with roughly 90% of his net worth given his ETH holdings, then asked whether Shapira held the same implied position on Ethereum. Shapira later posted that after Buterin's comments and those from Eliezer Yudkowsky, he had updated his confidence down from 50% to 40%.
What Shapira's AI security argument actually claims
Shapira did not specify a mechanism in his original post. The claim rests on a broad concern that AI development could expose or exploit weaknesses in Bitcoin's security model that holders currently assume are robust. That concern splits into at least two distinct categories: AI-assisted attacks on the network's cryptographic foundations, and AI-assisted exploitation of softer infrastructure like wallet software, node implementations, or mining pool operations.
Buterin's response addressed both without conflating them. He placed the probability of an actual cryptographic break, something that would require SHA-256 or Bitcoin's elliptic curve assumptions to fail, in a category he described as tiny. The network-layer category, covering client software and mining pools, he treated as manageable through the normal upgrade process that Bitcoin has used throughout its history without requiring social consensus across the full user base.
The distinction between cryptographic breaks and infrastructure attacks
SHA-256, the hash function underlying Bitcoin's proof-of-work system, has never been broken. The best-known theoretical attack against SHA-256 uses Grover's algorithm on a quantum computer, which would reduce the effective security from 256 bits to 128 bits. Security researchers generally consider 128-bit security sufficient against any foreseeable attack. NIST's post-quantum cryptography standards finalized in August 2024 did not treat SHA-256 as requiring replacement, focusing instead on public-key cryptography based on elliptic curves. Bitcoin's proof-of-work uses SHA-256 as a hash function rather than as a public-key system, which means the quantum threat relevant to Bitcoin's signature scheme, ECDSA on the secp256k1 curve, is separate from any threat to mining.
The more immediate AI-related concern in security research circles involves the use of large language models and automated vulnerability discovery tools to find and exploit bugs in software implementations rather than break underlying cryptography. Bitcoin Core, the dominant node implementation, has a public codebase and a security disclosure process. Historically, critical vulnerabilities in Bitcoin's implementation have included the 2018 CVE-2018-17144 inflation bug, which allowed double-spending under specific conditions and was patched before exploitation. The question Shapira's claim implicitly raises is whether AI-accelerated vulnerability research changes the rate at which such bugs are discovered and exploited before patches reach the network.
Why the transition problem is harder than the cryptographic problem
Buterin's framing of the primary challenge as "getting the transition" points to a well-documented problem in network security. Even when a cryptographic upgrade is technically straightforward, coordinating the entire user base of a decentralized network to adopt it creates a window during which some participants remain on older, potentially vulnerable software.
Bitcoin's upgrade history illustrates this. The shift from uncompressed to compressed public keys, the introduction of SegWit, and the Taproot upgrade each required years of coordination across miners, exchanges, wallet developers, and node operators. SegWit activated in August 2017 after nearly two years of debate, and adoption of SegWit-formatted transactions remained partial for years afterward. A post-quantum signature upgrade would be substantially more complex than any prior Bitcoin soft fork because it would require changing the fundamental cryptographic assumption that every existing address relies on, not just adding an optional new transaction format.
StarkWare researcher Avihu Levy demonstrated on August 26 that a quantum-resistant Bitcoin transaction could be constructed and mined on mainnet without a protocol change, using a technique called signature grinding that relies on hash functions rather than elliptic curve cryptography. StarkWare CEO Eli Ben-Sasson said at the time that a soft fork remains the better long-term answer, but the transaction showed that individual holders can take protective steps before any network-wide consensus is reached.
What Yudkowsky's contribution likely covered
Shapira credited both Buterin and AI safety researcher Eliezer Yudkowsky with moving his confidence from 50% to 40%, without specifying what Yudkowsky argued. Yudkowsky has written extensively about transformative AI risks but has also been skeptical of crypto specifically.
Yudkowsky and Buterin have disagreed publicly on AI risk timelines and severity before. Yudkowsky has argued that sufficiently advanced AI poses an existential threat on a shorter timeline than most mainstream AI researchers acknowledge, while Buterin has expressed more measured views about AI safety and has written about the complementary relationship between cryptography and AI safety mechanisms. Their apparent agreement in pushing Shapira's estimate downward on the Bitcoin security question suggests that whatever Yudkowsky contributed, it was skeptical of the AI-breaks-Bitcoin thesis rather than supportive of it.
Bitcoin traded near $79,500 at the time the exchange took place.
Tanya Petrusenko
Disclaimer: All materials on this site are for informational purposes only. None of the material should be interpreted as investment advice. Please note that, despite the nature of much of the material created and hosted on this website, HODL FM operates as a media and informational platform, not a provider of financial advisory services. The opinions of authors and other contributors are their own and should not be taken as financial advice. If you require advice, HODL FM strongly recommends contacting a qualified industry professional.
