A missing eight-byte check in an offchain code file handed an attacker a window to drain roughly $4.5 million from a Risk Labs-operated relayer on July 17, 2026 — yet every user of the Across protocol walked away without losing a cent. The Across Solana relayer attack exposed a narrow but consequential gap between what a blockchain sees and what offchain software chooses to trust.
Summary
Key takeaways
- On July 17, 2026, an attacker exploited a bug in Risk Labs’ offchain Solana relayer by forging deposit events that never existed on-chain.
- No user funds were lost or at risk; every genuine transfer was completed or refunded the same day.
- Risk Labs absorbed approximately $4.5 million gross in relayer capital, with a net loss under $4 million and shrinking as recovery continues.
- The attacker submitted 1,627 forged deposits across 18 destination chains totaling roughly $41.7 million in face value; the relayer filled 581 of them before Solana was disabled.
- Solana service was restored in approximately 12 hours via fallback CCTP routing; law enforcement and SEAL 911 are actively involved in recovery efforts.
Details of the Across Solana Relayer Attack
The incident unfolded inside a narrow technical seam between Solana’s onchain programs and the offchain software Risk Labs used to read them. No funds moved on-chain during the attack. No state changed. The blockchain itself behaved perfectly — but the software watching it did not.
Technical cause of the vulnerability
Solana’s SpokePool contract emits events using Anchor’s CPI-event pattern: the program invokes itself with event data through a designated event_authority PDA. The SvmCpiEventsClient component inside the Risk Labs relayer accepted any inner instruction addressed to the SpokePool through that PDA as a genuine event. The flaw was specific: it did not verify the 8-byte Anchor event discriminator that distinguishes real event CPIs from arbitrary instructions dressed up to look like them.
That omission gave the attacker a precise tool. They deployed a wrapper program that CPI’d into a benign, read-only SpokePool helper function — get_unsafe_deposit_id — and appended forged FundsDeposited payloads to it. From the relayer’s perspective, these looked like legitimate deposit events. Onchain, nothing happened. No funds moved, no balances changed, no contracts were touched. The deception existed entirely in the offchain reading layer.
Risk Labs later clarified that get_unsafe_deposit_id was not itself the vulnerability. It was a harmless read-only helper that the attacker simply used as a carrier. The root cause was the missing discriminator check in the relayer’s own code base.
Attack scale and forged deposits
Between 05:07 and 06:14 UTC on July 17, the attacker submitted 1,627 forged deposits from 1,627 single-use wallets, spreading them across 18 destination chains. The face value of those forged positions totaled approximately $41.7 million, all consolidating to a single EVM recipient address.
Risk Labs’ relayer filled 581 of those fraudulent requests, paying out roughly $4.5 million of its own capital before the team managed to disable Solana as an origin chain. The remaining approximately $37 million in forged deposits expired worthless once the relayer stopped filling them.
User Protection and Financial Impact
No user funds were lost at any point during the incident. That outcome was not luck — it was a structural consequence of how Across was built.
No user funds lost or at risk
Across operates as an intents protocol. Users deposit assets into an escrow contract on the origin chain; independent relayers advance their own capital to fulfill the transfer on the destination chain. The protocol only reimburses relayers after a separate settlement process verifies each deposit they filled. That ordering meant the forged deposits never touched user escrow. Every real user deposit was either completed or fully refunded the same day.
Equally important: no smart contracts were compromised. The Solana programs and all EVM contracts behaved exactly as designed throughout the incident. The entire vulnerability lived in offchain relayer software, not in any onchain logic.
Losses limited to Risk Labs’ relayer capital
Risk Labs absorbed the financial damage directly. The gross payout from the relayer was approximately $4.5 million. With roughly $500,000 in attacker funds trapped inside the protocol, the net loss sits under $4 million and continues to shrink as recovery efforts advance.
The contrast with traditional bridge architecture is worth pausing on. In a conventional lock-and-mint bridge, a vulnerability like this could give an attacker direct access to a shared pool of user funds. In an intents-based system, the relayer’s own capital absorbs the first loss — a design choice that, in this case, fully insulated every user.
Incident Response and Recovery
Timeline of key mitigation steps
The response moved quickly once the attack was identified. Here is the sequence of actions taken on July 17 (all times UTC):
- 05:07 — first forged deposit submitted on Solana
- 06:14 — forged deposit stream ends; first attacker addresses blacklisted at 06:16
- 08:23 — Solana disabled as an origin and destination chain in the API
- 08:35 — Solana SpokePool paused on-chain
- 08:36 — first public statement issued
- 09:37 — root-cause fix merged
- 10:26 — fix deployed across all Risk Labs infrastructure
- 17:05 — Solana deposits re-enabled via CCTP routing
The root-cause fix was deployed roughly five hours after detection. Full Solana service was restored in approximately 12 hours through fallback CCTP routing. All other protocol operations remained unaffected throughout.
Routing changes and system updates
Solana order flow is now routed exclusively through fallback CCTP routing, which Risk Labs says covers all major chains. Intents routing to and from Solana remains disabled while the team re-audits Solana offchain event parsing logic to confirm that this class of vulnerability is fully closed and that the security model around event handling is standardized across the stack.
More advanced anomaly detection techniques are also in progress, aimed at automatically flagging and halting this type of coordinated forged-deposit pattern before significant capital is paid out.
Law enforcement involvement
Recovery efforts extend beyond the protocol itself. Risk Labs is working with SEAL 911, which provided near-instant support after the incident was reported, and with US law enforcement. Attacker addresses have been flagged across exchanges and off-ramps. The ACX token buyout process, separately, remains completely unaffected and continues as planned.
Security Model and Lessons Learned
Design of Across’ intents protocol
The incident became a live test of a deliberate architectural bet. When Across was designed as an intents protocol, the decision to have relayers advance their own capital — rather than draw from shared user escrow — was a risk-management choice that many users may never have noticed. On July 17, that choice proved its worth. The attack’s financial impact was real, but it was contained entirely within Risk Labs’ own operational capital.
That said, the incident also illustrates a structural tension that any protocol relying on offchain event parsing must reckon with. Onchain contracts can be formally verified and audited with high confidence. Offchain relayer software operates in a grayer space — it interprets chain activity and acts on it, but the interpretation logic sits outside the trust guarantees of the blockchain itself. A single missing discriminator check in that layer was enough to generate $4.5 million in fraudulent payouts.
Future mitigation plans and audits
Risk Labs has committed to re-auditing its Solana offchain event parsing logic even while Solana order flow runs exclusively through CCTP — meaning the vulnerable path is already inactive. The goal is to ensure the vulnerability class is completely closed before intents routing to and from Solana is re-enabled, and to standardize how event security is handled across the broader infrastructure.
The deeper question the incident raises is how the industry treats offchain components of onchain protocols. Smart contract audits are now standard practice. Offchain relayer logic, event listeners, and parsing libraries have historically received less scrutiny — partly because they sit outside the formal security perimeter most auditors focus on. If the Across incident reshapes how protocols approach that gap, the $4 million loss may end up funding a broader security upgrade across the cross-chain settlement space.
FAQ
Did any user funds get lost in the Across Solana relayer attack?
No user funds were lost or at risk at any point. All genuine transfers were completed or fully refunded on the same day, July 17, 2026.
What was the technical cause of the security vulnerability in the Risk Labs relayer?
The vulnerability was caused by missing verification of the 8-byte Anchor event discriminator in the offchain relayer code. This allowed an attacker to forge deposit events that the relayer treated as real, even though no funds moved on-chain.
How did Risk Labs respond to the forgery attack on their Solana relayer?
Risk Labs disabled Solana as an origin chain, paused the Solana SpokePool on-chain, deployed a root-cause fix within approximately five hours of detection, and fully restored Solana service within roughly 12 hours using fallback CCTP routing.
Are smart contracts involved in the Risk Labs relayer attack compromised?
No. No smart contracts were exploited. All Solana programs and EVM contracts behaved exactly as designed throughout the incident. The vulnerability existed entirely within the offchain relayer software.
Article produced with the assistance of artificial intelligence and reviewed by the editorial team.
