Ethereum co-founder Vitalik Buterin rejected a Sept. 7 prediction that artificial intelligence could undermine confidence in Bitcoin’s security and cause its price to fall by more than 50% within two years.

Investor and AI-risk commentator Liron Shapira assigned a “50% confidence” level to his prediction that Bitcoin would suffer such a decline because AI could weaken what users consider the network’s security or robustness, according to his original post.

Buterin said he took “the opposite side” of that forecast. His response focused on Bitcoin’s technical adaptability rather than its market price, arguing that most AI-related threats could be addressed without changing the network’s core social agreement.

Buterin sees transition risk rather than cryptographic collapse

Buterin said he remains optimistic about cybersecurity over the long term. In his view, the main risk lies in managing the transition to a world where attackers and defenders both use increasingly capable AI systems.

He separated operational attacks from failures of Bitcoin’s underlying cryptography. Network-layer exploits, client vulnerabilities and attacks against mining infrastructure could require software updates, but they would not necessarily invalidate Bitcoin’s proof-of-work design.

Bitcoin’s network consists of several layers. Full-node software verifies transactions and blocks. Mining pools coordinate hash power, while miners provide the computing equipment that performs proof-of-work calculations. Internet infrastructure connects those participants.

AI could help attackers identify flaws in node software, wallet code, mining-pool systems or communications infrastructure. Defenders can also use AI to review code, test updates and detect suspicious behavior.

Buterin argued that client developers and mining pools could respond to those attacks through ordinary upgrades. Such changes would be less difficult than an emergency alteration of Bitcoin’s monetary rules or transaction history.

He described the probability of “actual breaks on hashes or PoW” as “tiny.” The statement is an assessment rather than a measurable guarantee, and it does not cover every cybersecurity risk affecting Bitcoin users.

Bitcoin’s proof of work is not the likeliest AI target

Bitcoin mining depends on SHA-256, the hash function used to process block headers and prove that miners performed computational work. A direct cryptographic break would require an attacker to find a practical weakness that sharply reduces the work needed to produce valid hashes.

However, Current generative AI systems have not demonstrated that capability. AI can improve software analysis and automate attack discovery, but it does not automatically defeat established cryptographic functions. An attacker also could not take control of Bitcoin simply by producing persuasive text, malicious code suggestions or automated social engineering. Compromising wallets, exchanges or individual developers would create losses, but it would not necessarily compromise Bitcoin’s consensus rules.

The more realistic near-term risks involve surrounding infrastructure. AI agents may improve phishing, credential theft, malware development and vulnerability discovery. Mining pools, exchanges, wallet providers and node operators could face more automated attacks.

A July security incident showed the distinction. Bitcoin Optech reported that a weakness in some COLDCARD-generated wallets produced insufficient entropy and exposed affected funds to theft. The problem involved wallet key generation rather than Bitcoin’s hash function or proof-of-work mechanism.

The technical report estimated losses exceeding 1,000 BTC at the time. Users of affected devices were advised to move funds to wallets created with secure external entropy.

Such incidents can damage confidence and cause major losses while leaving Bitcoin’s underlying ledger operational. They support Buterin’s argument that ecosystem software and custody remain more immediate attack surfaces than SHA-256.

AI could accelerate both attacks and defensive reviews

Shapira’s forecast appears to rest on an imbalance between attackers and defenders. Advanced AI could discover vulnerabilities faster than open-source developers can examine, patch and distribute software.

Bitcoin’s public codebase gives researchers broad access, but it also allows attackers to inspect the same software. A sufficiently capable automated system could review large repositories, identify unusual interactions and test attack paths continuously.

The Bitcoin Policy Institute and more than 40 digital-asset organizations raised that concern in August. They asked leading AI laboratories to provide vetted open-source security teams with controlled access to advanced models.

The coalition argued that defenders need early access to the same tools that sophisticated attackers may use. Its proposed program included computing resources, secure testing environments and direct communication with AI laboratory security teams.

Block, Coinbase, Strategy, MARA, Galaxy, BitGo, Brink, Trezor and several Bitcoin development groups supported the appeal. Crypto.news reported that the coalition sought frontier AI access for Bitcoin security researchers before more capable models became widely available.

A separate assessment from Bitcoin Policy Institute researcher Zack Shapiro and security specialist Efrat Fenigson’s coverage warned that the next several years could present a difficult transition. The concern was not that AI would directly solve Bitcoin’s proof of work, but that attackers could gain access to capable security tools before defenders.

As crypto.news previously reported, researchers warned that Bitcoin developers could fall behind AI-assisted attackers without comparable models, funding and computing access.

Social consensus would matter after a deeper failure

Buterin distinguished upgradeable operational problems from attacks that require social consensus. Bitcoin nodes and mining pools can adopt patched software when developers identify a conventional vulnerability.

A deeper cryptographic failure would be more difficult. If an attacker could forge digital signatures or bypass proof-of-work requirements, developers might need to introduce new cryptographic standards and coordinate a broad network migration.

Node operators, miners, exchanges, custodians and users would then need to agree on which software and transaction history to recognize. Disagreement could split the network or delay protective action.

Quantum computing is frequently discussed in this context because a sufficiently capable machine could threaten the elliptic-curve signatures protecting some Bitcoin holdings. That issue differs from Shapira’s AI claim because quantum computers use different computing methods rather than machine-learning techniques.

Crypto.news has reported that developers are exploring quantum-resistant infrastructure and migration tools before practical attacks emerge. Estimates about when such systems could threaten Bitcoin remain uncertain.

Buterin did not claim Bitcoin could respond easily to every cryptographic break. His argument was that those fundamental failures have a very low probability, while more plausible network attacks can be addressed through coordinated software changes.

Bitcoin showed no reaction tied directly to the debate

Bitcoin traded near $79,590 on Sept. 7, down approximately 0.5% during the latest session. Its intraday range was roughly $79,460 to $80,494.

There was no evidence connecting that movement to the exchange between Shapira and Buterin. Bitcoin’s price responds to liquidity, derivatives positioning, exchange-traded fund flows, macroeconomic data and broader risk sentiment.

Shapira’s prediction also does not specify a technical attack, affected software component or sequence connecting AI adoption to a 50% market decline. It is a probabilistic forecast and cannot be treated as a verified security assessment.

Buterin similarly offered no formal risk model supporting his confidence. His response provides a technical counterargument: Bitcoin can patch many operational vulnerabilities, while a direct break of its hash function or proof-of-work mechanism remains unlikely.

The dispute therefore concerns the speed and difficulty of Bitcoin’s security transition rather than evidence that AI has already compromised the network.

What we know about the victims

What we know about the carrier and the plane

A reported “white hat” actor has taken nearly 4,000 Bitcoin worth about $319 million from the Liquid Network, according to an incident update posted by the Blockstream-run sidechain community. Liquid subsequently paused bridge operations and asked exchanges to stop both LBTC deposits and withdrawals while it investigates what went wrong.

Liquid Network says the withdrawal was executed via SideSwap using a Peg-out Authorization Key, while insisting that the key used was not compromised. Still, the federation wallet balance shown in Liquid’s explorer dropped sharply—from roughly 4,200 BTC to about 207.275 BTC—prompting renewed scrutiny of how Liquid’s peg security functions when something unusual bypasses expected controls.

Key takeaways

• Liquid says bridge nodes were disabled temporarily, effectively pausing the Liquid sidechain until the issue is resolved.
• The incident involved an LBTC peg-out executed through SideSwap, with Liquid stating the Peg-out Authorization Key was not compromised.
• Liquid told exchanges to pause LBTC deposits and withdrawals while the team attempts to contact the actor and assess security gaps.
• An OP_RETURN message claimed the funds were extracted by “whitehats,” but neither the claim nor the technical details are fully verified publicly.

Liquid freezes bridge activity after a major LBTC outflow

According to the initial reporting in Liquid Network’s incident communications, a “shade under 4000 Bitcoin” worth approximately $319 million was withdrawn from Liquid. Liquid Network also referenced an unverified on-chain message—via OP_RETURN—asserting responsibility and asking to be contacted “on chain.”

In response, Liquid disabled bridge nodes, stating this stops any new transactions from being submitted to the network. The operational consequence is straightforward: without bridge nodes, the sidechain’s peg mechanics can’t continue normally, which is exactly what traders and exchanges need when a suspected peg-out route may be functioning unexpectedly.

“Bridge nodes have been temporarily disabled, so no new transactions can be submitted to the network. Effectively, the Liquid sidechain is paused until this issue is resolved.”

How Liquid’s peg-out is supposed to work—and what the incident challenges

Under standard Liquid mechanics, LBTC is burned on the sidechain before Bitcoin is released on the main chain. The withdrawal flow depends on authorization rules that require a multisignature setup (Liquid describes this as 11-of-15 multisig functionaries) and a whitelist for approvals.

That structure is meant to prevent exactly the kind of unauthorized peg-out that would drain funds from the federation wallet. The incident therefore raises questions that go beyond the size of the withdrawal: it challenges whether the controls around approvals and whitelisting performed as intended, or whether there is an unexpected pathway in the way approvals are generated and executed.

Crypto analyst DBCrypto argued that the behavior appears more consistent with an extraction that leaves funds “sitting on Bitcoin” rather than being rapidly mixed, describing it as potentially closer to “whitehat extraction than theft.” At the same time, DBCrypto said the broader security implications remain serious: either the required signatures and authorization logic were effectively satisfied, or the whitelist/control mechanisms designed to block such events did not hold.

SideSwap role and Liquid’s assertion about key security

Liquid said the withdrawn funds were sent via Sideswap, specifically through the SideSwap PAK (Peg-out Authorization Key). In Liquid’s statement, the PAK used in the transaction was not compromised, and it claimed that no other related keys were compromised either.

Liquid also reported that it had already established how the LBTC involved in the order was created—through a bug in Elements software. While the incident details in the public account focus on the peg-out authorization process and the status of the key, the Elements reference matters because it suggests the failure may have started earlier than the final Bitcoin withdrawal itself.

For market participants, the key implication is practical: if an Elements-level bug can affect how LBTC is created or approved for peg-out, then the operational risk isn’t confined to a single malicious transaction. Instead, it may require a broader review of how sidechain issuance and peg-out eligibility interact, and how those conditions are validated before bridge processing is allowed to resume.

What to watch as Liquid and related operators investigate

Liquid and its ecosystem appear to be working through a familiar incident sequence: identify which steps deviated from expected behavior, confirm whether any authorization keys were actually compromised, and determine what fixes or compensating controls are necessary before restarting bridge functions.

At the time of the provided coverage, Blockstream and Adam Back had not posted public updates on the incident timeline, but Samson Mow (Jan3 CEO) said “everyone is actively working to resolve this.” The immediate items for users and exchanges are likely straightforward—follow Liquid’s instructions to pause LBTC deposits and withdrawals until the bridge is re-enabled and the underlying security question is addressed.

Until Liquid publishes more technical detail on the peg-out authorization flow, the nature of the Elements bug, and why the multisig/whitelist protections were insufficient (or circumvented), the central uncertainty will remain the same: whether this was a one-off exploitation path or a systemic weakness that could reappear in other peg operations. Readers should watch for the moment bridge nodes return and for any concrete post-mortem describing exactly which authorization or validation step failed.

Documentary filmmaker Lance Oppenheim’s first fiction feature Primetime—premiering here at the Venice Film Festival—fills in some background pixels on the show’s rise and fall, focusing particularly on an ambitious, egotistical, and at least somewhat delusional Hansen, played by Robert Pattinson. It’s 2006, and Hansen and his beleaguered producer Andie (Merritt Wever, bringing an edgy nervousness to the role of a woman struggling to stay afloat in the ruthless world of TV news) get a chance to move their show into a primetime slot. The stakes are high, but the potential rewards are great. Hansen’s hubris balloons, and he infects others with his own inflated sense of the public service he claims to be providing. To complete its mission of reeling in the bad guys, the show employs youthful-looking decoys posing as underage kids. One of these is an elflike performer named Del (singer-songwriter Phoebe Bridgers) who can impersonate either a boy or a girl; she passes some tips along to a nervous newcomer who will come to be known as Decoy Dan (Skyler Gisondo, in a performance laced with touching eagerness). Though he hopes to become a real actor and thinks this is his way in, he wants to quit when he sees how demeaning the gig actually is. But Hansen, with his crocodile smile and purring, almost sultry intonation, persuades him to stick with it, assuring him that he’ll be helping to cleanse the world of evil.