Blockaid Says $287 Million Can't Be Frozen. The Real Number Is $1.09 Billion.
Start with the arithmetic, because the arithmetic is where this report breaks.
Blockaid's January-to-August 2026 exploit review counts 60 incidents and $1.18 billion stolen. Of that, $9.8 million came back to victims โ 0.83%. Another $81.4 million sits frozen but undistributed โ 6.9%. And $287.1 million remains in wallets nobody can freeze โ 24.3%.
Add those up. $379.7 million. Now subtract from $1.18 billion.
$802 million โ roughly 68% of all stolen value โ has no stated status anywhere in the report. Not frozen. Not returned. Not described as sitting in traceable wallets. Simply absent from the accounting.
That gap is the story. The headline everyone will quote โ $287 million that can't be frozen โ is a floor, and a flattering one. It describes money we can still see and cannot touch. The money we cannot see is 2.8x larger.
True unrecovered total: $1.0888 billion. A 92.3% attacker success rate, measured by value.
Context.
Blockaid builds pre-transaction simulation and screening tools. This is a vendor's self-published aggregation, relayed by Unchained, with no independent verification, no published address list for the 60 incidents, no inclusion criteria, and no taxonomy of attack vectors โ private key compromise, contract bug, governance capture, oracle manipulation, flash loan. That last omission matters most. Without vector classification you cannot tell whether risk is shifting structurally or merely scaling with TVL.
There is also a timestamp problem. The report covers "January to August 2026" with data cut off September 2. Cross-check: $229 million is tied to 107,165 ETH, implying roughly $2,137 per ETH. Run that against the calendar and something is off by a year. Transcription error, forward-looking content, or source problem โ verify the publication date before you cite any of this in an investment memo.
None of that invalidates the numbers. It tells you what kind of document you are holding: usable data, motivated framing.
Core.
The most valuable line in the report is not a dollar figure. It is a duration.
Median time from exploit to fund movement: 17 minutes. 68% within one hour. 22% within one minute.
For nearly a quarter of these 60 incidents, post-hoc defense was not difficult โ it was physically impossible. One-minute exfiltration does not happen opportunistically. It requires pre-signed transactions, pre-deployed drainer scripts, pre-mapped laundering routes. These were not crimes of opportunity. They were engineering projects with a launch button.
So the pivot from response to pre-execution blocking is technically correct. Note, carefully, that it is also Blockaid's product line. The logic holds. The logic also pays the vendor.
Now the structural finding, buried under the report's own headline.
Freezability is not a security capability. It is an asset architecture property.
Four layers, four answers. Stablecoin issuance โ freeze is trivial via issuer blacklist; the report cites zero cases. L2 bridge and governance โ freeze available via Security Council multisig upgrade; Arbitrum, $71 million, Kelp DAO. Chain consensus โ validator isolation or burn; Stellar, $6.9 million, Blend; Syscoin, $9 million, unauthorized token destruction. Native assets with no issuer โ ETH, BTC, wrapped tokens, LSTs โ freeze capacity is zero. No mechanism exists. $287.1 million.
That $287 million is not a failure of security. It is permissionless design working exactly as specified. ETH and BTC have no subject who can be compelled to execute a freeze order. That absence is the value proposition. It cannot be patched and should not be.
Which means the headline number is permanent and un-improvable. Framing it as an industry problem to solve is a category error.
And it is a thin sample dressed as a structural law: Drift alone accounts for $229 million โ 79.8% of all non-freezable funds. One protocol, one April incident, one number.
Red Flag: a 60-incident dataset in which a single case drives four-fifths of the central conclusion is not a dataset. It is an anecdote with a spreadsheet.
Contrarian.
Here is what recovery actually looks like once you stop aggregating.
Arbitrum's Security Council freeze of the Kelp DAO funds: $71 million, 77.9% of all recovered value. Syscoin's burn: $9 million, 9.9%. Stellar's validator isolation: $6.9 million, 7.6%. Everything else, every other mechanism across 60 incidents: $4.3 million, 4.7%.
Three events, 95.3% of all recovered capital. Five incidents out of 60 saw any freeze or return action at all.
That is not an industry capability. That is three protocols that happened to have a centralized lever within reach, and 55 that did not. Any "recovery is improving" narrative built on this dataset is sample selection wearing a lab coat.
Two more things the report gets wrong or leaves out.
Wrapped tokens are filed under "no central issuer, non-freezable." Wrong on its face. wBTC and wETH have custodians. The accurate statement is that the issuer did not respond to a freeze request โ a governance failure, not a technical impossibility. Conflating the two hides the only actionable lever available.
And not one stablecoin issuer freeze appears anywhere in the dataset, despite stablecoins being the most seizable asset class in existence. Read that as evidence of laundering-path arbitrage: attackers swap to ETH and BTC first, then bridge out of governance-reachable chains. Arbitrage is the market's way of telling you which assets can be clawed back.
Liquidity doesn't flee into assets it can be seized from. It flees into the ones it can't.
Takeaway.
The economics are settled and grim: 92.3% attacker success by value, 0.83% victim return, and a defense budget that must cover 100% of the attack surface to break even. No voluntary market clears that. Security demand in Web3 will be manufactured by regulators, insurers, and exchange listing gates โ or it will not exist.
Watch three things next quarter: whether Blockaid publishes its 60-incident address list, whether any stablecoin issuer freeze surfaces in the full-year data, and whether L2 governance tokens get repriced for the emergency powers their councils quietly hold.
Because if a Security Council can freeze $71 million with a multisig upgrade, then decentralization is a disclosure item, not a property. And nobody has priced it.