Unbacked mints: five 2026 incidents, one question the code forgot to ask

Five different bugs, one gap

Between 6 September and 1 October 2026, five projects lost money because their systems created something that looked like value and was not.

Liquid minted L-BTC no Bitcoin backed. NEAR Intents credited internal balances that no tokens backed. Meter's bridge minted wrapped MTRG with no collateral. Symbiosis's Bitcoin Bridge minted billions of syBTC from a 25-cent deposit. Payy's bridge released USDC against a burn that never happened.

The causes have nothing in common on the surface: a cache key, a failed callback, a validation flaw nobody has described in detail, an identity mix-up and a negative fee, and a proof verifier. What they share is one missing question, asked at the last possible moment: does the thing I am about to pay out against actually exist?

This piece is about that question. It uses each project's own account where one exists. We have full write-ups of Liquid, NEAR Intents, Meter, Symbiosis and Payy.

How each one got there

Liquid, 6 September. According to Blockstream's assessment, a cache that remembered the result of expensive proof checks could be tricked into skipping a check. The attacker's transaction created an output that its inputs did not back. Once accepted, the unbacked L-BTC looked like any other, and the normal peg-out process paid real BTC against it. Roughly 4,000 BTC left, and the assessment says about 602 BTC is still outstanding.

NEAR Intents, 29 September to 1 October. The postmortem says a feature that notified a third-party contract after a deposit could be made to fail on purpose. The token contract refunded the deposit but the ledger kept the credit. The ledger then promised more than the contract held. About $3.8 million left before the team paused the contract.

Meter, 23 September. Meter says a block validation flaw let an attacker mint unbacked wrapped tokens through its Passport bridge. About $2.3 million was estimated by Blockaid. There is no public technical post-mortem yet, so this one is the least understood.

Symbiosis, 11 September. According to its post-mortem, the bridge took depositor identity from data the sender controlled, which allowed the attacker to act as an administrator, set a negative fee and, through a second bug, inflate the credited amount. About 46.1 billion syBTC were created. Real losses were estimated at 9.97 BTC.

Payy, 24 September. Payy says an invalid burn proof was accepted by its verifier, and the bridge released about 1.92 million USDC.

The pattern

In every case, a component upstream said "yes", and the component downstream paid. The part of the system that moves value trusted the part that decides whether value exists.

That is natural. Splitting those jobs is how we keep systems modular. But it means the downstream part inherits every bug of the upstream part, and the upstream parts here are the hardest code in crypto: proof verifiers, caches of proof results, cross-chain message parsers and callbacks across contracts.

The five bugs also share a quality that makes them easy to miss in review. Each looked fine in isolation. A cache key joined its fields without separators. A callback failed in a way the ledger did not notice. A fee was just a number. Each only becomes dangerous in combination with something else the author did not think about.

What would have limited the damage

None of these controls fixes the underlying bug. They shrink the blast radius, and several of the teams have said they are adding them.

1. A solvency check that does not depend on the minting path. Compare total claims against total reserves, in a separate component, on every block or batch. NEAR Intents says it has now added such a check with an automatic pause. Liquid's own review notes that nothing compared L-BTC supply with BTC held before the peg-out.
2. Rate and size limits on exits. A cap on how much can leave per hour, and a delay on large withdrawals, buys time. Several of these drains took minutes.
3. A fast, tested pause. NEAR's contract could be paused within about 29 minutes of detection. Payy paused within the day. Some contracts, like Limit Break's old V2 and UNCX's vesting pool, had nothing to pause.
4. Independent verification for the highest-value paths. A single verifier is a single point of failure. Two implementations, or a second check on large payouts, make a forged proof a smaller event.
5. Keep proving systems and critical dependencies patched. In Payy's case, an analyst reports a fix had been public for 73 days before the attack.
6. Test the invariant, not just the function. Property-based and stateful fuzzing, with the rule "claims never exceed reserves" written down, finds this class of bug. It is boring, and it works.

What to watch

Meter's cause is still undisclosed, and Liquid's peg-outs remain suspended. If more detail arrives, the pattern may be sharper or messier than it looks today. We will update this piece if a project's own account changes the facts above.

Sources: Blockstream's assessment, the NEAR Intents postmortem, Symbiosis's post-mortem and Payy's incident page, as reported in our incident articles, plus Meter's statements via media. Only the official pages are linked.