Dev.to Security 🔐 Cybersecurity 👁 0 📖 4 min read

SwissBorg lost $41.5M through a partner API nobody could audit

SwissBorg lost about $41.5M in SOL in July 2026. The attacker never touched SwissBorg's own systems. They came in through a partner API. That single sentence explains most of the money lost in crypto this year, and a

SwissBorg lost $41.5M through a partner API nobody could audit

SwissBorg lost about $41.5M in SOL in July 2026. The attacker never touched SwissBorg's own systems.

They came in through a partner API.

Cartoon: a bank with a locked front door and an open side hatch labelled partner API, coins pouring out to a thief

That single sentence explains most of the money lost in crypto this year, and almost nobody's security review covers it.

What was hit

SwissBorg's "SOL Earn" staking product ran on top of a partner's API. That partner was compromised. Through it, the attacker pulled roughly $41.5M in SOL out of SwissBorg's position.

Context for the size: per Arkham's data, SwissBorg held about $72.6M in SOL. This was 57% of their entire SOL treasury in one incident.

The damage was contained in one way: only SOL Earn was affected, and roughly 1% of users were in that product. Other staking products were untouched. The CEO, Cyrus Fazel, went live the same day and committed the treasury to refunding a significant portion, with full reimbursement depending on how much gets recovered.

That is a fast, honest response. It is also what you say after the fact.

The part that should worry you

SwissBorg is not a sloppy operation. They have a security team, audits, monitoring. None of it mattered, because the thing that broke was not theirs.

You can audit your contracts a hundred times. If a vendor holds a key, or an API credential, or a withdrawal permission on your behalf, then your security is the weaker of your two systems. That is not a metaphor. It is literally the set of capabilities an attacker acquires.

And this is not an outlier. It is the current shape of the market.

The 2026 pattern

In the same window:

  • Kelp DAO: $292M, through a compromised LayerZero RPC node.
  • Drift Protocol: $285M, after a six-month social engineering campaign that ended in multisig keys.
  • Humanity Protocol: $36M, through a phishing email on an employee laptop that had production keys backed up to it.

Different entry points, same structural fact: the attacker did not beat the protocol's cryptography. They went around it, to whatever the protocol trusted.

The aggregate numbers make it clear. H1 2026 saw 182 incidents, up about 50% year over year, for roughly $956M in losses. Of that:

Category Losses
Wallet compromises $444M+
Bridges $328M+
Supply chain $298M+

Meanwhile, losses per protocol exploit fell from an average of about $156M in 2020 to 2022 down to roughly $14M from 2023 onward. Code got better. Audits got better. The economics of attacking a well-audited contract got worse.

So the money moved to the edges.

Why the edges are worse

A smart contract is finite. You can read it, test it, formalize parts of it, and reason about what it can do. That is why the audits eventually started working.

A trust graph is not finite. It is a list of everyone who can move your money: the staking partner, the RPC provider, the oracle, the bridge, the deployer, the engineer with a laptop, the vendor who has a support account, the API key in a config file.

Each one of those is a door. You can lock your front door perfectly and still lose everything through a window you forgot you installed.

The uncomfortable asymmetry is that you cannot audit the other side. You can ask a vendor for their SOC 2 report. You cannot verify that their internal credential handling matches it. You are trusting a document about a system you do not control, in an industry where the counterparty is often a two-person team with a good landing page.

What to actually check

If you run a product that touches staking, delegation, or any third-party custody, these four questions are worth an hour today:

1. Who is the backend partner, actually?
Not the brand on the marketplace page. The entity holding the keys that can move the funds. If you cannot name the company and the person, that is the answer.

2. What can that API do?
Read balances, or initiate withdrawals? Is there a cap per transaction, per day? A withdrawal limit is not bureaucracy, it is a circuit breaker. A permission that can move an unbounded amount is a single point of total loss.

3. Is there a time lock?
If a partner can drain a position instantly, you have no recovery window. If there is a 24-hour delay on large withdrawals, then the incident becomes an alert instead of a loss.

4. Is there a guarantee fund?
Not a promise, a funded reserve with a stated policy. "We would make users whole from treasury" is a sentence that depends entirely on how big the loss is relative to the treasury.

The line worth remembering

Your yield is not worth more than your principal.

Every staking product on the internet offers some percentage. That percentage is compensation for risk, and the risk is mostly not the smart contract anymore. It is the partner, the operator, the API, the key.

SwissBorg's users did nothing wrong. They picked a large, audited, public company. The failure happened one step to the left of everything they could see, in a system they were never shown.

That is the actual lesson of this year: the attack surface is not your code. It is your trust graph, and nobody publishes a picture of it.

The archive: 61 investigations, 64 leak scans, 0 exploited

The archive. 61 on-chain investigations and 64 days of GitHub leak scanning, redacted and never used. Read the index or try the free token risk check at qanzhi111.github.io.

📰 Read the original article on Dev.to Security

Originally published by Dev.to Security. Aggregated on AIWithGhost for educational purposes — full credit and traffic to the original publisher.