How Discount Bug Bounties Are Fueling Crypto's Billion-Dollar Loss Streak

A wave of 2025 exploits is exposing a simple pattern: protocols that skimp on bug bounties keep paying far more to attackers than they ever would have paid to the researchers who could have warned them first.

01A textbook failure at SuperRare

SuperRare's staking contract carried a flaw so elementary that commentators noted even an AI chatbot could have flagged it. The underlying issue was a straightforward access-control error — the kind that unit testing, or almost any bug bounty program, should have surfaced well before launch. SuperRare had no public bounty program running at the time. The result was a $730,000 exploit that experts described as easily preventable.

02The 2025 numbers

Industry-wide loss figures for 2025 make the SuperRare incident look small. The Chainalysis mid-year report put cumulative crypto losses at roughly $2 billion by the middle of the year — already surpassing the total for all of 2024. The pace didn't slow afterward: July losses reportedly hit $142 million, and August added another $163 million across 16 separate breaches.

03GMX and KiloEx: two exploits, two recovery bounties

Two mid-2025 incidents illustrate how the same protocols that under-invest in preventive security are still willing to pay after the fact. GMX's legacy V1 contracts were drained of roughly $42 million in July 2025 through a re-entrancy exploit — though public reporting doesn't clarify whether those legacy contracts were even covered by GMX's existing bounty program. Separately, KiloEx lost $7.5 million in what researchers labeled a "basic Oracle 101" style manipulation, the sort of issue proactive incentives are supposed to catch early.

Both protocols responded the same way: offering a 10% recovery bounty to get the funds back. GMX's attacker kept a $5 million reward for returning the rest. KiloEx's attacker took the offered $750,000 (10% of the loss) and sent the stolen funds back. In both cases, the money for a large reward clearly existed — it simply wasn't allocated until after the damage was done.

04What audits leave out

A large academic review of smart contract upgrade practices found further evidence that the industry's security gaps are structural rather than accidental. The study examined 83,085 upgraded contracts and catalogued 31,407 potential security issues, organized into eight risk categories — several of which the researchers said remain poorly understood or inadequately addressed across the industry. The same research pointed to a recurring problem: components frequently get marked "out of scope" during audits, only to become the entry point for real attacks later. The paper cites the exploited components in the Cork Protocol incident as an example of exactly this dynamic — audit boundaries that didn't match where the actual risk lived.

05The economics, according to Immunefi

Immunefi CEO Mitchell Amador has argued for some time that crypto security is fundamentally an incentive problem, not a code-quality one: disclosure needs to be made more profitable than exploitation, or rational actors will eventually choose exploitation. His recommended benchmark is a bounty worth about 10% of the funds at risk — meaning $10 million in protected assets should correspond to bounties as high as $1 million.

Instead, some platforms tie their cheapest service tiers to fixed bounty ceilings as low as $50,000, regardless of how much value is actually secured. Cork Protocol's public bounty, hosted on Cantina, was capped at $100,000 — yet the protocol suffered a roughly $12 million exploit on May 28, 2025, a loss 120 times larger than the maximum reward on offer. Amador has described the resulting dynamic as a race to the bottom that, in his words, creates "a death spiral that benefits no one except malicious actors."

06The counter-examples

Some protocols have taken the opposite approach, and their track record supports Amador's thesis. LayerZero launched a $15 million bug bounty program with Immunefi in 2023 — one of the largest such programs ever created — reflecting the scale of value the protocol moves across chains; it remains active today.

Euler Finance took a similar path after painful experience: following a $197 million flash loan exploit in 2023, the protocol expanded its original $1 million bounty program into a $7.5 million program on Cantina with broader scope than before.

MakerDAO set an early benchmark in 2022 with a $10 million bounty program. Now rebranded as Sky, the protocol still runs that program on Immunefi with the same $10 million ceiling, now covering the SKY governance system and the USDS stablecoin. Wormhole, meanwhile, paid out $10 million in 2022 for a single critical vulnerability report — the largest such payout on record at the time — and continues to run an active program with a current maximum reward of $5 million.

Immunefi reports that its platform has paid out more than $120 million in bounties overall while helping protect upward of $25 billion in user funds — evidence, in Amador's framing, that aligned incentives function as intended.

07A familiar mistake

The pattern echoes an earlier chapter in Web2 security history, where chronic underpayment and poor treatment of researchers drove many away from public bug bounty programs and toward less visible channels. Crypto now risks repeating that mistake at a moment when institutional capital is moving on-chain and regulatory scrutiny is intensifying.

SuperRare, Cork Protocol, GMX, and KiloEx differ in chain, vulnerability type, and attack vector, but the underlying failure is consistent: bounty budgets set well below the value actually at risk. Academic researchers have already catalogued more than 31,000 unresolved security issues in production contracts. Bounty platforms continue to report record payouts even as hack losses also hit record highs — a sign that, for now, the economics still favor attackers in far too many cases.