Allbridge Flash-Loan Exploit: How Pool Manipulation Drained $1.65 Million

The Allbridge incident wasn’t a black-swan mystery. It looked like a textbook pool manipulation play powered by a flash loan, executed quickly on Solana, and routed across chains before alarms could fully catch up.

We’re going to walk through what actually happened, how the attacker bent a stable pool out of shape, where the funds went, and what controls likely would have blunted it. No scare tactics. Just the mechanics and the fixes.

If you build, you’ll get a checklist. If you’re a user, you’ll see what to watch for the next time a bridge or pool starts behaving oddly.

Point Details Protocol pause and loss Allbridge Core paused on July 19, 2026 after about $1.65 million left its Solana pools, according to multiple security firms and industry reporting The Block. Attack path An attacker used a roughly $1.12 million USDC flash loan from Kamino, then cycled USDC↔USDT swaps to skew Allbridge pool ratios before withdrawals The Block. Cross-chain exit Funds were bridged off Solana to Ethereum shortly after the exploit, per alerts from PeckShield and CertiK Decrypt. On-chain traces Post-mortem traces show a 948,927.53 USDT withdrawal and a $2.24 million USDC movement through Allbridge as liquidity was shifted; TVL fell quickly after AMBCrypto. Primary weakness Stable pool pricing could be manipulated in a narrow window without enough dynamic fees, TWAP guards, or circuit breakers to resist a flash-loan burst.

What actually happened on Solana

Here’s the quick timeline as pieced together from on-chain traces and security alerts.

July 19, 2026: Pause and alerts

Allbridge Core paused the protocol after security firms flagged irregular activity that pointed to a loss of about $1.65 million from Solana liquidity pools. This pause was reported publicly the same day The Block.

Flash loan and swap spiral

On-chain analytics cited by The Block show the attacker took roughly $1.12 million in USDC through a flash loan on Kamino, then executed rapid USDC to USDT and back swaps to distort the stable pool’s ratio. This happened across July 19 to 20 as the positions cycled in and out within very tight blocks and transactions The Block.

Funds exit to Ethereum

Soon after the pool imbalance was monetized, security monitors PeckShield and CertiK flagged that the stolen funds moved out of Solana and onto Ethereum, consistent with a cross-chain laundering step Decrypt.

Post-incident traces add granularity: a 948,927.53 USDT withdrawal and a separate $2.24 million USDC movement were recorded as the liquidity migrated through Allbridge. The protocol’s TVL on Solana dropped sharply after the event AMBCrypto.

Inside the pool manipulation

Stable pools try to keep a near-1:1 price between assets like USDC and USDT with a curve that offers very low slippage around parity. That’s great for traders. It’s also a liability if an attacker can temporarily push the pool off-balance and settle a withdrawal at a price that doesn’t fully reflect external markets.

The basic recipe

1. Borrow a large amount of capital for a single block or short window using a flash loan.
2. Hammer the stable pool with swaps in one direction, tipping the ratio. USDC becomes scarce, USDT becomes plentiful, or vice versa.
3. Exploit the distorted price by withdrawing liquidity or swapping back at a favorable rate before the pool or price oracles adjust.
4. Repay the flash loan and pocket the difference.

All of this can settle within seconds if the chain is fast and the pool’s defenses are lenient.

Why the math breaks under stress

Curve-like invariants make trading near 1:1 cheap. Outside that band, slippage ramps up. If parameters like the amplification coefficient, fees, and rebalancing guards are set too loosely, a well-funded attacker can still shove the pool out of the safe zone and take advantage of a stale price path.

// Pseudocode thought experiment // If pool weights skew too quickly without fee escalation // effective_price deviates from external price for a few blocks if (skew > threshold && fee_multiplier == low) { attacker_profit = price_gap * size - fees; }

Defenses that help include time-weighted average price checks, dynamic fee multipliers that spike when liquidity tilts, and circuit breakers that stop swaps when an imbalance crosses a threshold.

Flash loans and execution on Solana

Flash loans compress step one. You don’t need capital on hand. You just borrow size, execute the plan, and pay back the loan within the same transaction or transaction bundle.

In this case, on-chain data points to a roughly $1.12 million USDC flash loan sourced from Kamino on Solana. That capital fueled the burst of USDC to USDT and back swaps that bent the Allbridge pool. The speed matters here. Solana’s throughput lets attackers run a high number of atomic or near-atomic actions in a tiny window, reducing the chance that slow-moving guards catch up The Block.

A word on MEV and sandwiching

When a pool is being pushed around, miners or validators, or on Solana, block producers and priority fee bidders, may insert or reorder transactions. That can worsen the outcome for defenders, but it can also provide signals for detection. Watching for clustered priority-fee spikes around specific pools is a practical early-warning indicator for operations teams.

Pro tip: If you run a pool on Solana, subscribe to mempool or block-builder data for large stable-pair flows. A sudden, tightly packed series of alternating swaps is a red flag worth rate-limiting in real time.

Where the money went

The immediate objective in pool manipulations is simple arbitrage of the mispriced pool. The follow-on objective is exit liquidity and obfuscation. Here’s what public traces suggest.

• Liquidity was yanked as soon as the price distorted, including a recorded withdrawal of 948,927.53 USDT, and a separate $2.24 million USDC movement through Allbridge as imbalanced liquidity migrated off the pool AMBCrypto.
• Soon after, the attacker bridged value from Solana to Ethereum, consistent with a common laundering step that spreads risk across chains and venues. This cross-chain move was flagged by PeckShield and CertiK in their public alerts Decrypt.

Once funds are on Ethereum, the set of mixers, DEXes, and OTC paths gets larger. That doesn’t make recovery impossible, but it fragments the chase for investigators. The earlier the freeze, the better the odds.

Controls that would have helped

No single control stops every flash-loan play. Stacking measures is the goal. Here is a practical map of controls to expected effects.

Control What it changes Trade-off Dynamic fees tied to imbalance Jacks up cost of pushing the pool off 1:1 in a short window. Can hurt user UX during volatility; needs tuning. TWAP checks on swaps and withdrawals Prevents settlement at prices far from recent averages. Lag can create stale quotes if windows are too long. Imbalance circuit breakers Pauses swaps if reserves skew beyond a threshold. Outages if sensitive; attackers may walk the threshold. Withdrawal throttles after large swaps Slows capital extraction long enough for rebalancing. Annoying for power users; needs clear UX. Oracle cross-checks for stable pairs Blocks trades if on-chain price diverges from trusted refs. Oracles can fail; must fail safe, not fail open. Pre-commit and batch settlement Reduces atomic manipulation by separating intent from fill. Complexity and latency; not ideal for all chains.

Pro tip: Do not rely on a single knob like fees. Attackers will simply budget for it if the rest of the system is predictable.

Impact on users and projects

TVL shock is the first-order effect. Even before official statements land, liquidity providers pull capital to avoid being the last ones out. That alone can deepen a pool’s imbalance and widen spreads.

Users typically want answers to three questions after a hit like this:

1. What funds left and via which assets? In this case, a mix of USDC and USDT with specific withdrawals and movements observed on-chain AMBCrypto.
2. Where did the funds go next? Crossed to Ethereum shortly after the event, per PeckShield and CertiK alerts Decrypt.
3. Will there be recovery or restitution? That depends on tracing success, negotiations if any, insurer terms, and treasury flexibility. It varies case by case.

If you’re an LP, assume mark-to-market loss until proven otherwise. If the project arranges a make-good plan, great. But set expectations based on the protocol’s treasury, insurer relationships, and how much of the funds can be frozen or clawed back.

Risk note: Even “safe” stable pools carry smart contract and market-making risks. A low-volatility strategy can still suffer capital impairment if defenses fail during an attack window.

Practical steps for teams and users

For protocol teams

• Simulate flash-loan attacks against your pool parameters. Fuzz the amplification, fees, and pause thresholds to see where manipulation becomes profitable.
• Deploy dynamic fees tied to reserve imbalance and recent flow volatility. Keep a kill-switch that routes to a safer mode, not a full halt, when possible.
• Add TWAP oracles for stable-pair pricing. Block large withdrawals if pool price drifts from external references beyond a tight band.
• Throttle or batch withdrawals following abnormal flow. Even a 60 to 120 second buffer can collapse an attacker’s window.
• Instrument real-time alerts for alternating-direction stable swaps, priority-fee spikes, and reserve skew. Give ops the tools to react without a governance vote.
• Run war games. Include an outside red team to test freeze, unfreeze, and comms cadence. Public confidence depends on how surgical your response looks.

For everyday users and LPs

• Watch pool ratios. If a USDC-USDT pool starts quoting weird prices or slippage jumps abruptly, step back. That’s often the first visible symptom.
• Check official channels for pauses. A protocol pause, like the one Allbridge initiated on July 19, signals the team is triaging. Don’t rush in for bargains unless you can price the risk The Block.
• Diversify stablecoin exposure across venues and chains. Concentration risk turns a small bug into a big drawdown for your portfolio.
• Prefer pools with documented circuit breakers and published param updates. If you can’t find docs on fees, TWAP, and pause logic, assume a softer target.

Pro tip: When in doubt, pull LP capital first and ask questions second. Re-adding liquidity is cheaper than eating an impairment that takes months to resolve.

Lessons for bridges and Solana builders

Bridges sit at the junction of market structure, consensus timing, and UX. That’s a lot of moving parts, and attackers pick on the weakest link. A few takeaways specific to Solana and cross-chain flows:

• Speed cuts both ways. Solana’s throughput lets users trade cheaply, but it gives adversaries a narrow, powerful window. Build for fast detection and fast throttling.
• Cross-chain exits are standard operating procedure for attackers. Your monitoring can’t end at chain boundaries. Set up coverage on the most likely exit chains, especially Ethereum.
• Parameter transparency helps. Publish how fees flex, how TWAP is computed, and what triggers a halt. Markets forgive incidents more easily than ambiguity.
• Incident drills matter. Pre-drafted comms, exchange contacts, and forensic partners save hours when minutes count.

If you want more grounded analysis like this, we cover security incidents and on-chain shifts without the noise at Crypto Daily. No hot takes. Just the facts and the implications.

Frequently Asked Questions

Was this purely a flash-loan attack or did a contract bug play a role?

From public reporting and on-chain traces, the core vector was pool manipulation powered by a flash loan and fast swaps. There’s no confirmed claim of a low-level contract bug being the primary cause in the public domain at the time of writing.

Why use Kamino for the flash loan?

Kamino is a Solana lending venue that can offer size and speed. The attacker needed quick access to seven-figure liquidity and a short settlement window. On-chain data points to roughly $1.12 million USDC sourced that way, which was enough to shove the pool around The Block.

How did the attacker profit if USDC and USDT are both stable?

By pushing the pool away from 1:1 temporarily, swap quotes become favorable on one side. The attacker could then withdraw or swap at an off-market rate before the pool repriced, repay the flash loan, and keep the spread.

What signals show funds moved to Ethereum?

Security firms PeckShield and CertiK published alerts noting cross-chain movements after the exploit. That aligns with common attacker playbooks of shifting value to deeper liquidity venues quickly Decrypt.

Can LPs expect reimbursement?

It depends. Some projects use insurance funds or negotiate partial returns if attackers respond to contact. Others may not have the balance sheet. Users should watch for official updates and avoid assumptions until a plan is published.

What would have stopped this specific attack?

A combination of dynamic fees when the reserve ratio skews, TWAP checks to reject off-market withdrawals, and an imbalance circuit breaker likely would have made the attack uneconomical or forced it to unwind early.

Is Solana uniquely vulnerable to this?

No. Flash-loan driven pool manipulation has hit multiple chains. Solana’s speed amplifies both attacker advantages and defender options. The difference is whether your controls operate at Solana speed.

Disclaimer: This article is provided for informational purposes only. It is not offered or intended to be used as legal, tax, investment, financial, or other advice.