Cross-chain transfers rely on an accounting promise: an asset issued or acknowledged on one network should correspond to an asset locked, burned or otherwise removed from circulation elsewhere. When that promise fails, the destination chain may release perfectly real funds against a claim that was never properly funded. The October 23, 2026 Initia incident and an August attack on The Sandbox’s SAND bridge reveal two distinct routes to the same economic failure. In one, attackers reportedly created unsupported account balances without changing the token’s published total supply; in the other, forged deposit messages led to unbacked tokens whose later burns authorized releases from a genuine vault. [1][2][3]
A Bridge Transfers Claims, Not the Same Coin Across Chains
A token cannot literally travel from one blockchain ledger to another. A lock-and-mint design places the original asset in custody on the source network and issues a corresponding representation on the destination. Reverse transfers burn that representation and release the original collateral. Burn-and-mint designs remove units on the origin chain and issue units on the destination, depending on issuer permissions and architecture. Liquidity networks instead pay users from inventory and later rebalance between providers. These are different accounting systems with different insolvency and execution risks. [4][6]
For a simple 1:1 lock-and-mint bridge, the governing invariant is verified source collateral >= outstanding redeemable destination claims. If a bridge has 10 million tokens in its vault and 10 million properly issued destination claims, it starts with full nominal coverage. If an attacker creates two million extra claims without adding collateral, the asset backing ratio falls to 10/12, or 83.3%. Nothing in that example requires an attacker to steal the original tokens first. The liability side expands, leaving the same asset stock supporting more claims.
A crucial caveat: a bridge with multiple token representations, transfers in flight, liquidity buffers, redemption queues and issuer controls cannot be analyzed safely using a single totalSupply field. The invariant must be calculated per asset, chain, bridge route and actual redeemable claim. Double-counted escrow, burned tokens and synthetic balances can invalidate a naive numerator or denominator.
The October 2026 Initia Incident: Balances Without Matching Debits
| Measure | Reported Observation | Interpretation |
| Unsupported nominal balances | ~$7.48m (Bitquery) | Face value, not realizable proceeds |
| Actual funds released | ~$2.73m (Bitquery) | Asset value extracted from reserves |
| Noble-linked USDC inventory | ~$1.09m to ~$63 | Reserve depletion, not global USDC supply |
On October 23, Initia stopped its layer-one network and rollups following a security incident. Its subsequent public account attributed the problem to a MoveVM script-cache flaw that allowed forged account state or balances for certain bridged assets, while reporting that the native INIT supply and relevant token total-supply counters were not increased. Initia described about $2.7 million in assets leaving through cross-chain routes before the halt. An independent Bitquery investigation traced approximately $7.48 million of nominally fabricated positions and about $2.73 million in real assets released from reserves. Those figures are not interchangeable: one measures face value of unsupported balances, the other value successfully extracted. [2][3][12]
Bitquery reported that the Noble-linked USDC reserve serving Initia declined from roughly $1.09 million to approximately $63 within 36 minutes, and that ordinary withdrawal attempts then failed. Its analysis shows the economic importance of settlement inventory. A user does not care whether the destination transfer was formatted properly if the bridge pays that withdrawal by exhausting coins held for legitimate users. The incident demonstrates that an invalid local balance can become a valid-looking external withdrawal obligation. The precise vulnerability and transaction accounting remain subject to the project’s final post-mortem. [2]
For the retail holder, the actionable distinction is between network accounting, token supply, and redemption entitlement. A wallet screen can display a balance that does not correspond to deliverable backing. A block explorer may show that a bridge released actual ETH or USDC to a recipient, yet that record says nothing about whether the claim was originally earned through a matching deposit. Investors who already hold legitimate tokens in the affected environment can face frozen withdrawals or impaired liquidity even if their own wallets were not attacked.
The Sandbox Incident: A Forged Origin and a Genuine Redemption
The Sandbox published a detailed post-mortem on August 27, 2026. The project’s SAND integration on Base and BNB Chain permitted an unintended path to change messaging verification permissions; the attacker then submitted messages representing Ethereum deposits that never happened. Destination-chain contracts created unbacked SAND. Some of those units were sold for WETH on Base; others were burned in the reverse direction, and the Ethereum vault released legitimate collateral. The report says the actual collateral released totaled 14,742,341.84 SAND, and the attacker’s successful economic capture across routes was approximately $987,000, against a broader modeled impact of around $1.50 million. [1]
The revealing detail is that the subsequent reverse-bridge transfers were not themselves fraudulent in their execution path. The submitted burns were real, and the established delivery infrastructure handled them. The logical error was upstream: the tokens being burned had never represented genuine locked deposits. As The Sandbox’s report puts it, a vault validating the burn event could not independently determine the provenance of the tokens it was redeeming. This is why an unrelated arbitrage bot also managed to redeem about 642,472 unbacked SAND it acquired on the market before the attacker’s withdrawal settled. [1]
No private keys held by The Sandbox were reportedly compromised; the failure concerned the delegation and verification configuration. This matters because “no stolen keys” is not the same as “no unauthorized release.” Security can break at the interface between a token application’s power and the messaging system’s trust assumptions, even if routine signatures and contract calls are technically valid.
Where the Balance-Check Logic Breaks
A conventional bridge monitor may verify that a burn occurred on Chain B and the corresponding asset was released on Chain A. That is necessary but insufficient. The harder question is whether the units burned on B originated from a verified earlier lock on A, and whether any special minting or account-state mutation changed that relationship. In an attack, successful accounting at the final transfer stage can conceal missing accounting at the original issuance stage.
Figure 1. A conceptual collateral/claim invariant. Figures are hypothetical, not drawn from a bridge.
The failure modes can be organized as a hierarchy: forged source events, compromised or misconfigured message attestation, flawed destination authorization, unauthorized token minting or synthetic account credits, replayed or duplicated claims, and inadequate global reserve reconciliation. Each layer needs independent controls. On-chain signatures do not verify real-world or cross-chain provenance by themselves; they authenticate statements made within a particular configured trust framework. [5][7]
In the ordinary case, there should be an asset-level ledger reconciling source deposits, source releases, destination issuance, destination burns, fees and claims still in transit. Controls should reject duplicate identifiers, verify supported token mappings, restrict authority to alter verification settings, and impose route-specific withdrawal caps. Chainlink’s documentation explicitly discusses token pools, lock/release versus burn/mint designs and rate limits that narrow the maximum loss per route over time. Those mechanisms may limit damage, although they cannot replace a correct issuance invariant. [4][5][13]
Quantifying the Gap Between Fake Claims and Real Loss
The Initia case offers a stark example of the difference between headline token value and actual realized extraction. Using Bitquery’s approximate figures, $2.73 million of real withdrawals divided by $7.48 million in fabricated nominal positions equals 36.5%. That does not mean 36.5% of an honest reserve was always available, or that 63.5% of the fake positions retained a market value. It simply means the amount documented as escaping across these routes was about 36.5% of the face value constructed in the attacker’s account state. [2]
Figure 2. Initia values reported by Bitquery on October 23, 2026; not additive measures.
The Sandbox report provides a different calculation. Its reverse-bridge vault had 14,743,464.21 SAND immediately before the attacker’s carefully sized withdrawal, and the attacker minted 14,743,364.21 destination tokens for that operation—100 fewer than the vault balance. Other parties also redeemed some unbacked units. The eventual 14,742,341.84 SAND released from collateral represented 99.992% of that reported starting vault balance. This is not an estimate of all SAND ever minted on the compromised networks, and it is not a statement that 99.992% of all investors’ assets were lost. [1]
It is also why “the original chain’s native token supply did not change” is a narrow assurance. The Sandbox said that its Ethereum token supply stayed fixed at three billion even while the bridge vault lost inventory and unbacked SAND proliferated elsewhere. Initia similarly reported no change to INIT’s native issuance. Investors must ask which balance or supply has been measured, for whose benefit, and against which redemption promise.
How a Bridge Should Prove That Each Claim Was Funded
| Control | What It Proves | What It Cannot Prove Alone |
| Authenticated message | Accepted verifier signed payload | Underlying deposit actually occurred |
| Collateral reconciliation | Claims covered by verified assets | Future exploit impossible |
| Rate limit | Bounds maximum pace of releases | Each claim genuinely backed |
For each withdrawal, an independent reconciler should link the destination token’s issuance lineage to a finalized authorized deposit, verify uniqueness of the source message, and subtract prior redemptions. The control should also compare aggregate liabilities to actually available collateral, not merely check a message’s cryptographic validity. This is conceptually adjacent to why exchange proof-of-reserves reports can be insufficient: verifying an asset balance without capturing the complete obligation set is not proof of solvency.
The bridge’s design should be understood before funds are deposited. A user holding a native issuer token differs from a holder of a wrapped or bridged claim. The distinction also appears in our analysis of asset representations traded as DeFi tokens, where equivalent price exposure does not entail equivalent legal rights or redemption processes. For bridges, the relevant questions are custody, issuer, route, mint permissions, halt powers, finality and compensation policy.
The immediate technical response after an incident is often to suspend affected routes, cap flows, identify fraudulent balances, preserve transaction traces and reconcile entitlements before reopening. The Sandbox said its affected integrations could not safely reopen because their vulnerable contracts were immutable; a replacement architecture would separate the token and messaging application. That is a structural response rather than a promise that the original code can simply be “patched” in place. [1]
What the Latest Incident Tells Retail Users
Read our report on Initia’s October 2026 minting incident alongside this mechanical analysis: the central issue is not that a bridge invented money in the abstract, but that unsupported claims were exchanged for real assets held on behalf of other users. A legitimate withdrawal may fail once reserves are depleted even if a user’s tokens and private keys are uncompromised.
Before bridging, identify the exact chain pair and whether the destination asset is native, canonically issued, escrow-backed or supported by third-party liquidity. A low fee or attractive quote cannot compensate for opaque verification and collateral accounting. During an exploit, do not buy severely discounted wrapped tokens on the assumption that they remain redeemable at par. A visible price, an authentic transfer event and a correctly signed message are each incomplete evidence of backing.
What Would Falsify the Core Risk Argument?
A bridge could materially reduce this class of risk by proving, for every redeemable unit, a finalized and unique source-chain deposit and showing continuous independently verifiable reconciliation. Formal verification, independent monitors, constrained mint authority, route caps and deliberately separated administrative permissions would make unbacked claims harder to create and harder to monetize. A permissionless implementation also needs credible governance controls, not merely a good audit report. This is a framework for evaluating architecture, not a claim that any particular commercial bridge eliminates all failure modes.
A post-mortem that establishes a different source of funds, a different root cause or a changed user-compensation outcome should revise the case assessment. Incident reporting can evolve over days. As of October 24, Initia had discussed a patch and investigation; its final technical post-mortem was not yet a verified source for the analysis. The Sandbox’s August post-mortem was more mature, providing transaction-level calculations and explicit bridge-replacement plans.
Why a Valid Message Can Still Authorize a Bad Payment
A cross-chain verifier answers a bounded question: did an authorized set of participants attest to this message under the current configuration? It does not necessarily independently calculate the issuer’s global assets and liabilities, prove that a token’s local balance was authorized by all intended ledger invariants, or inspect an earlier transaction for semantic correctness. If an application can change who is trusted, the verifier may work exactly as specified while the policy it enforces becomes dangerously permissive. The Sandbox report’s delegation failure belongs in this category. [1][7]
This explains why a security audit focused only on signatures or message replay protection can miss asset-coverage failures. A complete review needs to test token mint permissions, account-state transitions, privilege delegation, collateral custody, escape hatches, rate limits and governance upgrades as an integrated mechanism. A message may be correctly delivered, a burn event correctly logged and the vault release correctly executed, while the economic origin of that burn is nonexistent.
Liquidity Bridges Have a Different Kind of Liability
Not every bridge promises that one wrapped token is secured by one token frozen in a single vault. Liquidity-based bridges pre-fund destinations and earn fees for routing payments, effectively creating inventory and rebalancing exposure. The relevant safety metric is then available payout capacity across chains, liabilities owed to liquidity providers, settlement finality and how much outstanding credit is permitted. Using the simple 1:1 coverage formula without recognizing such a model can wrongly label a functioning bridge insolvent or miss a separate liquidity shortfall.
For users, the distinction affects the remedy if something breaks. A bridge-issued token holder may have a claim on issuer or escrow, whereas a customer using a liquidity router may primarily depend on route fulfillment and the provider’s contractual promises. A bridge warning cannot reasonably say “funds are safe” without specifying the particular token, chain, custody model and time. Cross-chain technology reduces the friction of moving exposure but does not erase underlying balance-sheet relationships. [4][6]
Rate Limits Change the Maximum Damage, Not the Proof of Backing
As an illustrative example, suppose an otherwise vulnerable destination can mint or redeem no more than 50,000 tokens per hour. A two-hour detection and suspension interval would bound cumulative released units to approximately 100,000 under a strict non-burst rule, compared with an uncapped path able to empty an entire large vault almost immediately. Real token-bucket limiters may allow an initial burst and refill over time, so a proper ceiling calculation must include starting capacity and the refill rate. Chainlink describes both values explicitly. [5]
Figure 3. Illustrative cumulative claim release with and without strict throughput limits.
The benefit is loss containment, not authentication of the source deposit. A false claim under the limit remains false; it simply cannot travel as quickly. Likewise, a multisignature is only as valuable as the independent control of its participants and the set of operations it actually authorizes. The most resilient design uses independent checks for the origin of assets, controls over issuance and narrow limits on how much damage an undetected mistake can cause.
Methodology and Limitations
The illustrated 83.3% coverage is $10 million assets divided by $12 million claims. The Initia 36.5% conversion ratio uses approximately $2.73 million divided by $7.48 million; the latter is a nominal valuation of fabricated balances and should not be read as recoverable market value. The Sandbox collateral-release ratio uses 14,742,341.84 divided by 14,743,464.21 SAND. All example fee and protection logic is conceptual, not a measured loss rate for blockchain bridges generally. Documentation reviewed through October 24, 2026.
Sources
Johan Shamshad is a financial markets writer at Dave Finances covering cryptocurrencies, trading platforms, brokers, fintech, financial regulation, and developments across global markets. He previously worked at Gulf News, adding newsroom experience to his coverage of fast-moving financial and digital-asset markets.
His work focuses on identifying market-moving events, company developments, regulatory changes, product launches, and shifts in trading and financial infrastructure.
Johan contributes news and analysis designed to help readers understand not only what happened, but why a development matters and how it may affect the wider financial landscape. You can reach out to him via his social media accounts:
Linkedin: https://www.linkedin.com/in/johan-shamshad-742851262/
X: https://x.com/Yasmine_FX
Investing: https://www.investing.com/members/contributors/279781574