Web3 security losses totaled $1.32 billion across 344 incidents in H1 2026, per CertiK's Hack3d report published July 6. The headline figure represents a 46.8% year-over-year decline from H1 2025's $2.47 billion. Strip out the $1.45 billion Bybit hack that dominated the prior-year period, and the...
"A headline reading of 'losses down nearly 50%' would suggest a meaningfully safer ecosystem. The data does not support that conclusion." — CertiK, Hack3d H1 2026 Report
Web3 security losses totaled $1.32 billion across 344 incidents in H1 2026, per CertiK's Hack3d report published July 6. The headline figure represents a 46.8% year-over-year decline from H1 2025's $2.47 billion. Strip out the $1.45 billion Bybit hack that dominated the prior-year period, and the adjusted comparison reverses: H1 2026 losses are approximately 28% higher on a like-for-like basis.
Incident volume more than doubled, rising from 83 in H1 2025 to 207 in H1 2026. The attack surface has shifted from smart contract exploits to human-layer compromises — wallet takeovers and targeted phishing now account for over 60% of dollar losses. Two North Korean state-sponsored operations alone drained $577 million, or 44% of all H1 losses, in an 18-day window. The G7 has formally designated these thefts a weapons-financing threat, though the June 2026 Évian statement included no new sanctions, no enforcement timeline, and no specific operational measures.
The 46.8% year-over-year decline in Web3 losses is almost entirely a function of base effects. H1 2025 was defined by the Bybit breach in February 2025 — at $1.45 billion, the largest single crypto exploit in history. Remove that event and H1 2025 losses stood at approximately $1.03 billion.
Against that adjusted baseline, H1 2026's $1.32 billion represents a 28% increase. As TRM Labs stated in its concurrent analysis: "The lower total reflects the absence of another record-setting theft, not a reduction in attacker capability."
The incident count tells a starker story. At 207 documented events, H1 2026 recorded the highest six-month frequency on record, per TRM Labs data. More attacks, more often, with individually larger payoffs per successful operation outside the top two mega-hacks.
Q1 2026 saw $508.2 million in losses, driven primarily by phishing. Broad-volume phishing campaigns declined in count — incidents fell 52.3% from 132 in H1 2025 to 63 in H1 2026 — but per-incident yields rose sharply. Losses declined only 10.8% despite the halved volume, indicating a decisive shift toward fewer, higher-value targets. Four phishing incidents accounted for approximately 85% of all phishing losses in the half.
Q2 2026 escalated to $807.5 million, a 59% quarter-over-quarter increase. Wallet compromises replaced phishing as the dominant attack vector. The KelpDAO exploit ($292 million, April 18) and the Drift Protocol breach ($285 million, April 1) together represented over 70% of Q2 losses. Both were attributed to North Korean state-sponsored actors.
CertiK's data breaks H1 2026 losses into three primary categories:
| Attack Vector | Losses | Incidents | Avg. Loss per Incident | |---|---|---|---| | Wallet Compromises | $444.5M | 33 | $13.5M | | Phishing | $366.3M | 63 | $5.8M | | Code Vulnerabilities | $151.6M | 204 | $743K |
The distribution contains a structural message. Smart contract exploits — the attack class that dominates security audit budgets and media coverage — accounted for just 11.5% of dollar losses despite representing 60% of incidents (125 of 207 total). The median smart contract exploit yielded sub-$1 million returns.
Wallet compromises, by contrast, produced 33.7% of all losses from just 16% of incidents. The average wallet compromise returned $13.5 million per event — 18 times the average code vulnerability payout. This gap reflects a fundamental misallocation of defensive resources across the industry: protocols spend heavily on code audits while leaving operational security, key management, and human-layer defenses underfunded.
Ethereum absorbed the highest incident count at 153 events, with $522.8 million in losses — 39.6% of the H1 total. The figure reflects Ethereum's continued dominance as the primary DeFi settlement layer and, consequently, the largest target surface.
Solana recorded only 7 incidents but $315 million in losses, driven almost entirely by the Drift Protocol exploit. The per-incident severity — $45 million average — was the highest of any major chain. Binance Smart Chain logged over 100 incidents but with individually smaller losses, reflecting its higher proportion of lower-value targets.
The Solana figure illustrates concentration risk: a single protocol compromise on a chain with fewer independent security layers can produce disproportionate damage relative to chain-level TVL.
North Korean state-sponsored actors have become the single most consequential threat to Web3 capital preservation. TRM Labs data shows a steady escalation in DPRK's share of global crypto theft:
| Year | North Korean Share of Global Crypto Theft | |---|---| | 2020-2021 | Under 10% | | 2022 | 22% | | 2023 | 37% | | 2024 | 39% | | 2025 | 64% | | 2026 YTD (through April) | 76% |
The two April 2026 operations were not smart contract exploits. They were multi-month human intelligence operations.
Drift Protocol (April 1): The Lazarus Group's TraderTraitor subgroup spent approximately six months building a front company posing as a crypto trading firm. Operatives attended industry events, held in-person meetings, and established credibility with Drift engineering staff. Once trust was established, malware was delivered to engineering endpoints, enabling credential and signing key access. The protocol's Security Council operated on a 2/5 multisignature configuration with zero timelock — a governance structure that, once compromised, permitted 31 unauthorized withdrawals totaling $285 million in approximately 12 minutes.
KelpDAO (April 18): Attackers compromised two RPC nodes servicing a LayerZero bridge and executed a DDoS attack to force failover, tricking LayerZero's single-verifier DVN into approving a fraudulent cross-chain transaction. The 116,500 rsETH stolen ($292 million) was partially frozen — approximately $75 million on Arbitrum — with roughly $175 million laundered through THORChain into Bitcoin before interdiction was possible.
Combined, these two attacks drained $577 million in 18 days through two structurally different vectors — social engineering governance signers and poisoning infrastructure RPCs. Cumulative North Korean crypto theft since 2017 now exceeds $6.75 billion, according to G7 estimates.
The G7 summit in Évian-les-Bains on June 17, 2026, produced a formal statement designating North Korean crypto theft as a weapons-financing threat. The leaders of Canada, France, Germany, Italy, Japan, the United Kingdom, and the United States stated: "We reiterate the need to jointly address North Korea's cryptocurrency thefts and cybercrimes."
The statement rests on three rhetorical pillars: enhanced policy coordination among member nations, stronger enforcement of existing sanctions frameworks, and disruption of laundering networks that convert stolen crypto into usable funds.
However, the Évian communiqué named no new sanctions, proposed no exchange regulations, specified no crypto mixer controls, and set no enforcement timeline. No specific operational measures were announced against wallets, platforms, or intermediaries linked to laundered funds. The gap between designation and enforcement remains wide.
The practical challenge is structural: THORChain and other cross-chain protocols processed $175 million in KelpDAO proceeds before any freeze was possible. Permissionless infrastructure, by design, does not accommodate the enforcement timelines that state-level responses require.
Of the $1.32 billion in H1 losses, approximately $115.3 million was frozen or recovered — an 8.7% recovery rate. Adjusted for recoveries, net losses stood at approximately $1.20 billion.
The low recovery rate reflects two factors. First, laundering speeds have accelerated: cross-chain bridges and decentralized exchanges allow stolen funds to be converted and dispersed within hours, often before freeze orders can be coordinated. Second, the jurisdictional fragmentation of crypto infrastructure makes coordinated interdiction operationally difficult even when funds are traceable.
CertiK recommended three defensive priorities: hardening private key management layers, implementing hardware-backed multisignature governance, and geographically distributing signing authority. These measures address the wallet compromise vector directly, though they do not mitigate the social engineering campaigns that precede most high-value attacks.
The H1 2026 data exposes a misalignment between where the industry spends on security and where losses actually occur. Smart contract auditing — a $500 million+ annual market — addresses a threat class that produced 11.5% of dollar losses. Operational security, key management, and anti-social-engineering training — the domains where 62% of losses originate — receive a fraction of that spend.
The Drift Protocol breach is the clearest illustration. The protocol had undergone multiple code audits. Its smart contracts were not the failure point. The compromise was a six-month social engineering campaign targeting human operators, followed by exploitation of a governance configuration (2/5 multisig, zero timelock) that no code audit would have flagged.
This is not an argument against smart contract auditing. Code vulnerabilities remain the highest-frequency attack class. It is an argument that the current security investment profile is mismatched to the actual loss distribution.
The Web3 security environment in H1 2026 has not improved. It has reorganized. The threat has shifted from exploiting code to exploiting people and infrastructure dependencies. State-sponsored actors now dominate the loss landscape, operating with intelligence-agency tradecraft — months-long infiltration campaigns, front companies, in-person relationship building — against an industry whose security model was designed to defend against anonymous remote attackers probing smart contracts.
The policy response lags the operational reality. The G7 has named the threat but not the remedy. Cross-chain infrastructure launders stolen funds faster than any interstate enforcement mechanism can freeze them. And the industry's own security spending remains concentrated on the attack class that produces the smallest share of losses.
The implication is not that code auditing should decrease. It is that operational security — key management, governance configurations, social engineering resistance, and infrastructure dependency mapping — must scale to match the threat that the data actually describes.