Dragonfly Estimates Prevention of Coldcard Hack at $2

cryptonews.ruPubblicato 2026-08-05Pubblicato ultima volta 2026-08-05

Introduzione

The venture capital firm Dragonfly estimates that the vulnerability exploited in the Coldcard hardware wallet hack could have been prevented for approximately $2 by using advanced AI to scan the code. Managing partner Haseeb Qureshi argued that cybersecurity is increasingly a matter of expenditure: with AI models conducting attacks, a product's security depends on how much the developer spends on scanning their own code with similar frontier AI tools compared to the attacker's budget. He based this estimate on the time an AI model took to find the bug and the API costs from a provider like Zhipu AI. Meanwhile, research from Galaxy indicates the attack was carried out by at least 15 different entities, with confirmed losses now estimated at $100 million and potentially rising to $130 million. The hack, which began on July 31, initially resulted in the theft of nearly 595 BTC from around 500 users. In response, hardware wallet companies Trezor and Foundation have warned users about related phishing campaigns.

An attack on Coldcard hardware wallets could have been prevented by checking the code with artificial intelligence for $2. This was stated by Dragonfly venture fund managing partner Haseeb Qureshi.

People are not appreciating one of the biggest takeaways from the COLDCARD hack.

Cybersecurity is now all about spend. Once AIs are doing all of the attacking, the simple question is how much money are you spending with frontier AIs scanning for vulnerabilities, compared to what... https://t.co/jkljyYtkxr

— Haseeb >|< (@hosseeb) August 3, 2026

Cybersecurity has boiled down to a question of expenditure, the expert believes. Vulnerability discovery on the attackers' side is now handled by AI models, so a product's security is determined by how much the developer spends on scanning its own code with similar tools — compared to the spending of the attacking side.

According to Qureshi's logic, the speed at which a neural network discovers a known bug serves as an indirect indicator of the cost to preemptively fix it. He applied this thesis to Coinkite, based on reports of Claude's eight-minute success with Coldcard. However, the Dragonfly partner considered the result dubious — the model could have used a published description of the problem.

In response, another user ran a test on $GLM 5.2 without internet access: the time increased to 20 minutes. Qureshi converted this into a monetary equivalent based on the API rates of Zhipu AI (the developer of $GLM): $1.4 per million input tokens and $4.4 per million output tokens. He tasked the Opus model with the calculation, which estimated the total cost at approximately $2.

"A $2 AI-powered security boost would have caught the [Coldcard] vulnerability," concluded Qureshi.

Losses Grow to $100 Million

According to Galaxy Research, at least 15 different malicious actors exploited the vulnerability. Analysts came to this conclusion after processing new reports from victims — unlike a hack of a centralized exchange where the scale is immediately visible, the picture here formed gradually.

now NUMEROUS different attackers exploiting the Coldcard vulnerability. we estimate at least 15 different attackers now

we continue to receive victim reports and give them info to report to authorities

and those reports help us identify new attacks and label attackers https://t.co/6ybBTqJPb6

— Alex Thorn (@intangiblecoins) August 4, 2026

"Thanks to one victim reporting a theft of less than 1 $BTC, we discovered a new attack that withdrew 12 $BTC from 126 addresses," wrote the firm's Head of Research, Alex Thorn.

Galaxy Research estimated losses from the three confirmed waves at $100 million. Including a suspected fourth wave, the total could rise to approximately $130 million.

Recall that on the night of July 31, approximately 500 Coldcard owners had 594.48 $BTC stolen. After the news broke, holders began transferring bitcoins to new addresses, not to exchanges, Glassnode analysts noted.

On August 4, Trezor and Foundation warned users about phishing campaigns in light of the incident.

Sleep at Night Technology: How Coldcard Turned Its Users' Sleep into a Nightmare
end-content

Domande pertinenti

QAccording to Haseeb Qureshi from Dragonfly, what could have prevented the Coldcard hack for approximately $2?

AHe stated that using frontier AI models to scan the code for vulnerabilities could have prevented the hack for approximately $2.

QWhat is Haseeb Qureshi's core argument about modern cybersecurity?

AHe argues that cybersecurity is now primarily about spending, specifically how much a developer spends on scanning their own code with advanced AI compared to what attackers spend on AI to find vulnerabilities.

QBased on Galaxy Research, what is the estimated total financial loss from the Coldcard hack?

AGalaxy Research estimates the losses from three confirmed attack waves to be $100 million, which could potentially rise to around $130 million with a suspected fourth wave.

QHow many different attackers did Galaxy Research identify as exploiting the Coldcard vulnerability?

AGalaxy Research identified at least 15 different attackers exploiting the Coldcard vulnerability.

QWhat action did many Coldcard holders take after the hack was publicized, according to Glassnode analysts?

AAccording to Glassnode analysts, many Coldcard holders transferred their bitcoins to new addresses rather than to exchanges.

Letture associate

Besu Patches Vulnerabilities in 5 Components: What Node Operators Need to Know

The Ethereum client Besu, developed by the Hyperledger community, has patched five security vulnerabilities discovered by blockchain security firm CertiK. These vulnerabilities, detailed in four security advisories on August 14, were all addressed in version 26.7.1, an urgent security update initially released on July 27. The intentional delay between the patch release and the public disclosure of details gave node operators a crucial window to update. JiaLiang Chang, CertiK's Director of Security Engineering, explained this "patch first, details later" model provides defenders a time advantage, allowing them to identify affected systems, test the update, and coordinate deployments—particularly important for institutional or permissioned blockchain networks requiring formal change management. The vulnerabilities, found through CertiK's "Chain Scan" attack methodology, involved issues in block announcement handling, consensus proposal buffering, WebSocket subscription limits, and JSON-RPC filter creation. If exploited, they could have allowed an attacker to exhaust a node's memory or thread resources, compromising its availability and the consensus process. Chang highlighted that while the open-source ecosystem is moving toward more formalized security testing (like differential fuzzing and bug bounty programs), coverage remains uneven. Testing often focuses more on protocol compliance than on continuous resource exhaustion, race conditions, or deployment-specific failures. He emphasized that third-party research remains vital for uncovering attack vectors beyond routine development, advocating for a mature, cumulative security model combining continuous integration, multi-node attack testing, independent audits, and regression testing for each confirmed vulnerability.

cryptonews.ru20 min fa

Besu Patches Vulnerabilities in 5 Components: What Node Operators Need to Know

cryptonews.ru20 min fa

Hack of Term Finance: Attacker Withdraws $8.5 Million Through Governance System Vulnerability

Decentralized lending protocol Term Finance lost $8.5 million due to an exploit of its governance mechanism, not a smart contract hack. The attacker manipulated the low voter turnout in the Meta Vaults system. By depositing just 0.5 ETH, they received a token (tmvETH) representing a share in the vault. Most users did not convert this token into a separate voting token (gtmvETH). The attacker did, gaining control of 90.66% of the issued voting power despite owning only 0.017% of the vault's capital. The attacker created a legitimate governance proposal to disable the 7-day withdrawal timelock and add their own contract to receive user funds. The proposal, publicly viewable for about 145 hours, passed due to a quorum requirement of only 5% and simple majority rule. After voting ended on August 23, 2026, the proposal was executed, allowing the immediate theft of 2,841 WETH and 1.68 million USDC (later swapped for DAI) from several vaults. The incident highlights systemic risks in decentralized governance when voting rights are separated from economic stake and low participation thresholds exist. Security mechanisms like timelocks were ineffective as the attacker could disable them from within the proposal itself. In response, Term Labs irrevocably closed all Meta Vaults and revoked DAO roles. The core lending protocol remained unaffected. The attack underscores the need for governance designs where security parameters are protected from modification through ordinary proposals.

cryptonews.ru20 min fa

Hack of Term Finance: Attacker Withdraws $8.5 Million Through Governance System Vulnerability

cryptonews.ru20 min fa

Trading

Spot
活动图片