Qubic Halts Chain After Weak Seeds Drain 1,404 Wallets in 2026

Qubic, the layer-1 network that pitches itself as a fee-less chain for AI compute, has been frozen since early Sunday after an attacker emptied 1,404 wallets of roughly 2.94 trillion QU, worth about $1.3 million. The blockchain itself was not broken. The keys were. According to the project, the wallet seeds behind those accounts were generated with insufficient randomness, which let the attacker rebuild them and sign transfers as if they were the owners.

That makes this more than a small-cap token story. It is the second high-profile weak-seed theft of 2026, after the Coldcard firmware incident disclosed in July, and it lands the same weekend Cysic shut its bridge after an unauthorized 10 million CYS mint. For anyone holding crypto in self-custody, the useful question is not whether they own QU. It is where their own seed phrase came from.

What happened on the Qubic network

The timeline is unusually precise because the chain stopped moving right after the theft. Crypto Briefing and Bitcoin.com News, both citing Qubic, put the drain between 23:42 UTC on Saturday, October 17 and 01:26 UTC on Sunday, October 18, a window of about 1 hour and 44 minutes. Qubic's validators, which the network calls Computors, then coordinated a halt at tick 85,371,949. A tick is Qubic's unit of network progress, roughly comparable to a block elsewhere.

Cointelegraph reported on Monday that the network was still paused at that tick, and Qubic has not announced a restart date. The key numbers so far:

• Wallets drained: 1,404, per Qubic. A community list behind the qubicmap.org checker matches that count and totals 2,939,606,531,944 QU, according to trenchbook, which traced the funds on Qubic's public RPC.
• Dollar value: about $1.3 million at the time, per Bitcoin.com and Crypto Briefing. The Defiant, using CoinGecko's price at 11:03 a.m. ET on October 18, put it at about $1.26 million, and stressed that this is a market valuation, not what the attacker could actually realize.
• Funds still on-chain: 87.1% of the stolen QU never left the Qubic network and is immobilized by the halt.
• Funds at an exchange: the remaining 12.9% went to MEXC, whose team froze the receiving account within about an hour, according to Qubic.
• Exchange response: Bitget separately suspended QUBIC deposits and withdrawals.
• Market reaction: Bitcoin.com reported QU was down more than 10% on Sunday morning. Cointelegraph put Qubic's market capitalization at about $62.6 million at the time of its report.

Trenchbook's on-chain tracing adds some useful texture. Of the seven wallets that collected the stolen QU, five had never sent anything out and held 2.07 trillion QU between them. One collector forwarded 379.04 billion QU, the 12.9% slice, to an address that passed it on within minutes, the pattern typical of an exchange deposit address. Qubic says that address belongs to MEXC; outside observers cannot see the exchange's side of the freeze.

The bug: Math.random() where a cryptographic generator belonged

Qubic's own framing is blunt: "The issue was in seed generation, not in the Qubic protocol itself." The project says the affected seeds are those created:

• in any browser on an iPhone or iPad before October 18, 2026,
• in Safari on a Mac before October 18, 2026, or
• in the Qubic iOS app before July 2025.

Seeds created elsewhere are not on Qubic's list. The fix shipped in web wallet version 4.26.4, whose release notes are dated October 18 and list a change to "use cryptographically secure seed generation," according to The Defiant's reporting syndicated on Cryptonews.net. Cointelegraph reviewed the associated GitHub commit: it replaces JavaScript's Math.random() with crypto.getRandomValues() in the wallet's seed-creation routines.

That one-line difference is the whole story. Mozilla's MDN documentation for Math.random() states that it does not provide cryptographically secure random numbers and should not be used for anything security-related; developers are pointed to the Web Crypto API's getRandomValues() instead. Math.random() is built to be fast and statistically uniform for things like animations or games, not to be unpredictable to an adversary. When its internal state can be narrowed down, the "random" outputs that follow can be reproduced.

Bitcoin.com described the effect in practical terms: Qubic seeds are 55 characters long, which should make guessing impossible, but the attacker apparently did not need to search that full space. Instead they reconstructed a much smaller set of predictable seeds and checked which ones controlled funded wallets. The fact that the affected list is tied to specific browsers and Apple platforms suggests the weakness depended on how particular JavaScript engines implemented the non-cryptographic generator, though Qubic has said a full technical report will only come after a recovery plan is settled. Until then, the exact amount of entropy the bad seeds had is not public.

Two more details matter for holders. First, Qubic says its team secured additional vulnerable funds before the attacker reached them. Second, QEARN, the network's staking feature, was reported as unaffected.

Halting a chain to stop a theft cuts both ways

The halt is the reason 87.1% of the stolen QU is still sitting where the attacker left it. On most large networks, a thief who has valid signatures is final the moment the transactions confirm. Qubic's Computors were able to stop the ledger cold, and the project now says the core team is identifying affected addresses and preparing recovery options for a Computor vote. Trenchbook reported that a community poll is open on Discord, and Qubic runs a separate Computor proposal and voting portal.

That is good news for victims, but it is a reminder of what kind of network this is. A validator set that can coordinate a full stop, and then vote on what happens to specific balances, is closer to a governed ledger than to an immutable one. Every holder of QU, not just the 1,404 victims, currently cannot move funds. Bitget's suspension and the MEXC freeze mean the token's main exit routes are also constrained.

This is not Qubic's first incident at the edges of its stack, either. Crypto Briefing noted that the Vottun Bridge connected to Qubic was exploited in August 2026. And Bitcoin.com recalled that the project, founded by NXT and IOTA veteran Sergey Ivancheglo ("Come-from-Beyond"), drew criticism in 2025 after claiming majority control of Monero's mining hashrate. None of that history caused this weekend's theft, but it shapes how much benefit of the doubt the market extends while the chain stays paused.

What to watch next, per the reporting so far:

• The Computor vote on whether and how the frozen QU is returned.
• A restart date for the mainnet on the patched client.
• Qubic's technical report on exactly how weak the seeds were.
• Bitget reopening QUBIC deposits and withdrawals, and how MEXC handles the frozen account.

A pattern, not a one-off: weak seeds in 2026

Cointelegraph placed Qubic in a longer line of attacks on wallet randomness: similar flaws hit Trust Wallet's browser extension and the Libbitcoin Explorer tool in 2023, and in July 2026 Coldcard disclosed that attackers had stolen funds from wallets whose seeds were weakened by a firmware bug. We covered that hardware case in our Coldcard firmware breakdown.

Coinkite's own advisory is worth rereading next to the Qubic news, because the two incidents share the same mechanics even though one is a dedicated hardware signer and the other a web wallet:

• The device or app was not "hacked" in the usual sense. Coinkite says Coldcard units were not remotely accessed or taken over; attackers regenerated private keys offline from weak seeds. Qubic says its protocol was not at fault.
• Entropy was the failure point. Coinkite says seeds from Mk4, Q and Mk5 units before the fixed firmware carried about 72 bits of entropy rather than the expected 128 bits, and Mk2/Mk3 seeds from firmware 4.0.1 (March 2021) through 4.1.9 were at greater risk.
• Updating does not fix an existing seed. Both projects shipped patches for new seeds. Coinkite is explicit that new firmware "does not change or repair an existing seed." Qubic's guidance, as relayed by trenchbook, is to treat an affected seed as weak even if it was not drained, and to move funds off it under Qubic's instructions once the network restarts.
• User-supplied entropy and passphrases helped. Coinkite says seeds created with at least 50 independent, private dice rolls are not at risk from the bug alone, and a strong, unique BIP-39 passphrase adds an independent barrier.

The broader security picture is not improving. Cointelegraph cited AMLBot's 2026 investigations report, which found hacks and exploits made up 19% of crypto crime cases but 40% of reported losses in the third quarter. Weak-seed incidents are a small slice of that, but they are among the hardest for users to detect, because nothing looks wrong until the balance is gone.

The same weekend: Cysic's bridge mint

Qubic was not the only project to freeze something over the weekend. The Block reported that an attacker used Cysic's cross-chain bridge to mint 10 million CYS on BNB Smart Chain at 8:31 a.m. ET on Saturday. On-chain records reviewed by The Block show the attacker sold 7.2 million of those tokens, bridged the proceeds off BNB Chain and received about 306.6 ETH (roughly $727,000) on Ethereum through Mayan, deBridge and SWFT's Bridgers before Cysic blacklisted the receiving address at 9:47 a.m. ET, freezing the remaining 2.8 million CYS.

The Cysic Foundation said the attacker "gained unauthorized access to a database used by services supporting the bridge and exploited a vulnerability in the relevant processing logic to submit a fraudulent mint message." It said the mainnet was not affected and that "existing user assets were not stolen." The bridge stays closed until a security review is complete.

The two incidents are different in kind. Cysic's was an infrastructure compromise in bridge back-end services; Qubic's was a cryptography mistake in wallet software. What they share is that the core chain logic held up while an adjacent component failed, and in both cases a centralized lever, a validator halt or a token-contract blacklist, limited the damage. Those levers protect users after the fact. They also confirm that the projects can freeze balances when they decide to.

What this means for investors

The Qubic drain is small in dollar terms, and Bitcoin was trading near $82,500 on Monday according to CoinDesk. For QU holders, the near-term risk is concentrated: funds are locked until the Computors vote and the chain restarts, and exchange access is limited. For everyone else, the lesson is about provenance of keys rather than price.

A few practical takeaways for self-custody users:

• Know where every seed was generated. If a seed came from a browser page or a web wallet, you are trusting that page's random number generator. Dedicated hardware wallets reduce that exposure but, as Coldcard showed, do not eliminate it. Our Ledger vs Trezor comparison and guide to the best Bitcoin wallets cover the main options and how they handle seed backups.
• A patch protects new seeds only. When a wallet vendor discloses an entropy bug, the safe response is to create a fresh seed on fixed software and move funds, carefully, with a small test transaction first.
• Add independent protection. A strong, unique BIP-39 passphrase, or user-supplied entropy where the device supports it, gives an extra layer that a flawed generator alone cannot defeat, according to Coinkite's guidance.
• Checking does not require your seed. Trenchbook noted that confirming whether a Qubic wallet was hit only requires the public address. Anyone asking for a seed phrase, or offering paid "recovery," after an incident like this is running the next scam; Qubic says its team will never message first or ask for a seed.

The deeper takeaway is about where risk now sits. Base-layer consensus failures are rare on established chains. The losses keep coming from the layers around them: wallet code, bridges, signing infrastructure and back-end databases. Qubic and Cysic both stopped the bleeding within hours, but only because each had a way to freeze what the attacker took. Investors should price in both sides of that: faster incident response, and a network that can be paused.