Crypto protocols lost $1.32 billion across 344 on-chain incidents in the first half of 2026, according to CertiK's Hack3D report. The headline figure represents a nominal 47% decline from H1 2025's $2.47 billion, but that comparison is misleading: strip out the single $1.45 billion Bybit breach f...
"April was the worst month in four years with only three days without a hack." — Ronghui Gu, CEO and Co-founder, CertiK
Crypto protocols lost $1.32 billion across 344 on-chain incidents in the first half of 2026, according to CertiK's Hack3D report. The headline figure represents a nominal 47% decline from H1 2025's $2.47 billion, but that comparison is misleading: strip out the single $1.45 billion Bybit breach from February 2025 and the comparable H1 2025 base drops to approximately $1.03 billion, putting 2026 roughly 28% higher on an adjusted basis.
The pattern has shifted. Fewer incidents now produce larger individual losses. Nearly 44% of H1 2026 losses originated from just two events — the $292 million Kelp DAO bridge exploit and the $285 million Drift Protocol governance hijack. Neither was caused by a smart contract bug. Both exploited operational security failures: compromised keys, social engineering, and infrastructure single points of failure. The August 2026 Coldcard hardware wallet exploit, which drained $130 million in bitcoin via a five-year-old firmware flaw, reinforced the same thesis. The attack surface has migrated from code to people and processes.
North Korea's Lazarus Group accounted for an estimated 66% of all crypto theft in H1 2026. The group executed the two largest exploits of the year and demonstrated operational patience measured in months, not days. The insurance gap remains wide: Nexus Mutual, the largest on-chain cover provider, has paid out $18 million in total lifetime claims — roughly 1.4% of H1 2026 losses alone.
CertiK's Hack3D report documents $1,315,676,432 stolen across 344 on-chain incidents from January through June 2026. The quarterly breakdown:
| Period | Incidents | Losses | Dominant Vector | |--------|-----------|--------|-----------------| | Q1 2026 | 44 | $482M | Social engineering | | Q2 2026 | 67 | $764M | Operational failures (88%) | | April 2026 alone | ~30 | $631M | Bridge + governance exploits |
April accounted for 68% of all first-half losses. The month recorded only three days without a reported exploit, according to CertiK. Attack frequency rose 68% year-over-year: 47 DeFi-specific incidents in the first 4.5 months of 2026 versus 28 in the same window of 2025.
By late July, year-to-date losses had crossed $1.3 billion across 276-plus incidents, according to Forbes. With the Coldcard exploit in early August adding another $130 million, the running 2026 total exceeds $1.4 billion through mid-August.
The composition of losses has changed materially. CertiK CEO Ronghui Gu stated that "the weakest link has moved from code to keys and people." Wallet compromise is now the costliest attack vector. The two largest H1 2026 incidents — Kelp DAO and Drift — both began with infrastructure compromise and social engineering, not vulnerability exploitation in audited smart contracts.
This represents a structural challenge for the security audit industry. The global DeFi security market was valued at $4.8 billion in 2025 and is projected to reach $22.6 billion by 2034. More than 80 identifiable security vendors compete globally. A standard mid-complexity DeFi audit costs $60,000 to $120,000 and takes 10 to 20 business days. But the two largest exploits of 2026 would not have been prevented by code audits at any price.
The implication is that protocol security spending, currently concentrated on smart contract review, is misallocated relative to where losses actually occur. Operational security — key management, multisig governance, personnel vetting, infrastructure redundancy — generates the majority of dollar losses but receives a minority of security budgets.
The largest single exploit of 2026 targeted Kelp DAO's rsETH bridge, built on LayerZero's cross-chain messaging protocol. According to Chainalysis and the LayerZero post-incident report, the attack began on March 6, 2026, when an attacker socially engineered a LayerZero Labs developer to harvest session keys and gain access to LayerZero's RPC cloud environment.
The root cause was a 1-of-1 verifier configuration: a single node was responsible for checking cross-chain messages before releasing funds. The attacker compromised internal RPC nodes and DDoS'd external nodes to feed false data to the verification network, authorizing a phantom token burn on the source chain and releasing 116,500 rsETH on Ethereum. Of the stolen funds, 89,567 rsETH was deposited on Aave as collateral to borrow $190 million in WETH.
LayerZero Labs stated it "made a mistake" in approving the configuration. Post-incident, the protocol committed to migrating all defaults to 5-of-5 Decentralized Verifier Networks (DVN) where possible, and no less than 3-of-3. Kelp shifted its bridge to Chainlink CCIP. Solv Protocol moved more than $700 million in tokenized bitcoin infrastructure away from LayerZero.
The second-largest exploit involved a six-month social engineering campaign by North Korean operatives targeting Drift Protocol, a Solana-based DEX. According to TRM Labs, the attack was attributed to TraderTraitor, a DPRK-linked threat group.
The attackers used Solana's "durable nonces" feature to get Drift Security Council members to unknowingly pre-sign transactions that transferred admin control. Once in control, they whitelisted a worthless fabricated token (CVT) as collateral, deposited 500 million units, and withdrew $285 million in USDC, SOL, and ETH — all within 12 minutes of execution.
Blockchain analytics firms Elliptic, Chainalysis, and TRM Labs independently confirmed DPRK attribution.
On July 30, 2026, attackers began exploiting a firmware flaw in Coinkite's Coldcard hardware wallet that had existed since a March 2021 release. A build configuration error caused seed generation to fall back on a weak software random number generator instead of the device's hardware entropy source, collapsing effective key strength from 128 bits to as low as 40 bits on older devices.
According to TRM Labs and Galaxy Research, approximately 1,816 BTC — over $130 million — was drained from more than 5,200 addresses. The core theft unfolded in roughly 25 minutes. No physical access to devices was required.
Bitcoin developer James O'Beirne reportedly raised the flawed randomness code with Coinkite in May 2025, more than 14 months before exploitation. He was told the issue would likely have surfaced already if it were real. As of mid-August, most stolen funds remain pooled at attacker-controlled addresses, with limited laundering activity: a single 64.9 BTC Wasabi deposit and 200 ETH deposited to Tornado Cash.
Cross-chain bridges remain the highest-risk infrastructure category in crypto. Eight major bridge attacks from February through mid-May 2026 produced approximately $329 million in losses. Combined with the Kelp DAO exploit, bridge-related losses exceeded $620 million in H1 2026.
Bridges combine multi-chain logic, off-chain relayers, consensus validation, and large asset custody into a single attack surface. Two cross-chain bridges — AFX Trade and Verus — were hacked on the same day (July 22-23, 2026) for a combined $31.5 million. Attackers gained access to private keys from five AFX bridge validators and withdrew $24.15 million in USDC.
The market response has been a migration toward more heavily validated messaging protocols. Chainlink CCIP, now live on 60-plus chains, implements a separate Risk Management Network running on an independent codebase to validate every message. Institutional users including Swift, UBS, and J.P. Morgan use CCIP in production for tokenized asset settlement.
DPRK-linked actors represented 66.2% of all crypto hacking losses in H1 2026, according to multiple blockchain analytics firms. In April alone, North Korea was responsible for an estimated 95% of that month's losses, executing 12 Lazarus Group-attributed attacks that yielded $635 million.
The group's cumulative all-time crypto theft exceeds $6.75 billion since 2017. The 2026 operations demonstrate increasing sophistication: the Drift attack involved six months of in-person social engineering before execution. The Kelp DAO breach required infiltrating a protocol developer's cloud infrastructure weeks before the actual drain.
The concentration of losses in a single state actor has policy implications. Standard protocol security measures — audits, bug bounties, formal verification — are designed to find code-level vulnerabilities. They are structurally inadequate against patient, well-resourced intelligence operations targeting human operators and infrastructure providers.
A new attack surface emerged in 2026: AI agents with wallet access. Over $45 million in security incidents involved autonomous AI trading agents. The first documented prompt injection exploit of a live wallet occurred when an attacker tricked an AI agent into transferring $204,000. In a separate incident, a prompt injection hidden in Morse code caused Grok to approve a transfer of 3 billion DRB tokens worth $174,000.
CertiK's Gu noted that the April exploit surge "could only be possible with AI," referencing the speed and coordination of attacks. He estimated an attacker can spend "$10,000 to $20,000 worth of compute tokens to keep advanced engines running continuous vulnerability scans against a protocol for days or weeks on end."
According to a MetaMask security report, 45.6% of teams relied on shared API keys for AI agents, making it difficult to trace or stop rogue actions. The asymmetry is structural: attackers use AI to find vulnerabilities faster, while defenders operate within fixed audit budgets.
On-chain insurance coverage remains a fraction of exposure. Nexus Mutual, the largest decentralized cover protocol, reports $5.2 billion in crypto assets safeguarded and $18 million-plus in total lifetime claims paid. That $18 million represents roughly 1.4% of H1 2026 losses.
Coverage in 2026 extends to smart contract exploits, stablecoin depegging, validator slashing, bridge risk, oracle risk, and liquidity provision risk. Premiums start under 1% annually. But the coverage gap reflects both supply constraints (limited capital pool for underwriting) and demand-side issues (protocols and users underinsuring relative to actual risk).
No bridge protocol has achieved complete security coverage. Insurance protocols analyze code quality, audit history, and total value locked to set premiums, but the shift toward operational and social engineering attacks makes traditional risk modeling less predictive.
The economics of protocol security are structurally unfavorable for defenders. CertiK reported more than 5,000 clients, each with a fixed security budget. A pre-launch audit for a mid-complexity DeFi protocol costs $60,000 to $120,000. Enterprise multi-chain systems exceed $150,000. With over 80 competing audit vendors globally, price competition creates race-to-the-bottom dynamics.
Meanwhile, Gu noted that "when [protocols] move assets on-chain, they need to face all these AI attacks, smart contract vulnerabilities, oracle manipulation, and cross-chain bridge hacks." The defender must secure every surface. The attacker needs to find one gap.
This asymmetry has institutional implications. CoinDesk reported that DeFi vulnerabilities remain traditional finance's biggest blocker to on-chain asset migration. The security gap is not abstract — it is a measurable cost that reprices the economic value of on-chain infrastructure for any institution conducting a total-cost-of-ownership analysis.
The data from H1 2026 describes a security environment where the nature of threats has outpaced the structure of defenses. Smart contract audits, the industry's primary security expenditure, would not have prevented the year's three largest exploits. The attack surface has migrated to key management, personnel vetting, infrastructure redundancy, and — increasingly — AI-integrated systems.
The concentration of losses in state-sponsored actors and operational failures suggests that incremental improvements to code auditing will produce diminishing returns. Protocols that survive this environment will need to redirect security spending toward operational resilience: multisig governance reform, infrastructure decentralization, real-time monitoring, and institutional-grade key management.
For institutional capital evaluating on-chain deployment, the calculus is straightforward. The total cost of on-chain operations must include not just gas fees and protocol costs but also the actuarial cost of security exposure — currently running at an annualized rate exceeding $2.8 billion industry-wide. Until the insurance market, operational security standards, and defense tooling close that gap, the security deficit will continue to function as a hidden tax on every dollar of value flowing through decentralized infrastructure.