Security

The security model, stated plainly

A security claim is only useful if it names the attacker and the damage. So instead of asserting that the agent model is safe, this page walks the scenarios that actually worry people and says what each one can and cannot produce.

In short

In the agent model, the worst case is hostile or incompetent trading inside your own account — bounded by per-leader notional ceilings, leverage caps and sub-account isolation. Withdrawals and transfers stay exclusively signable by your wallet, so no compromise of HyperMirror, its agent key or its infrastructure can move your funds to another address. The residual risks are market risk, execution risk and your own key security.

At a glance

Threat scenarios and their bounded consequences
Scenario	What becomes possible	What stays impossible
Platform or agent key compromised	What becomes possibleHostile trading inside caps and isolated sub-accounts	What stays impossibleWithdrawing or transferring your funds
Platform offline or shut down	What becomes possibleMirroring stops; positions stay open and unmanaged	What stays impossibleAny loss of access to your own funds
Leader account collapses	What becomes possibleThat sleeve's positions lose; emergency removal applies	What stays impossibleContagion into other leaders' sub-accounts through netting
Your wallet compromised	What becomes possibleFull account control by the attacker, including withdrawal	What stays impossibleNothing — this is why wallet hygiene matters most

Scenario: HyperMirror is compromised

Assume an attacker fully controls the platform and its agent key. They can submit orders for accounts that approved an agent, which means they can trade badly and cost money. They cannot withdraw to their own address, because an agent signature cannot express a withdrawal.

The bound on damage is structural rather than procedural: caps limit position size and leverage, and each leader's sleeve is a separate sub-account, so the exposure is not a single unbounded book.

Scenario: HyperMirror disappears

Mirroring stops because no new orders are being signed. Your funds are unaffected because they were never held anywhere else, and your open positions remain in your account under your key. There is no withdrawal queue to join and no claim to file.

The practical action in that scenario is to revoke the agent and decide whether to hold or close whatever is open.

Scenario: your wallet is compromised

This is the one case the architecture cannot help with, and it is worth being direct about it. A compromised wallet can withdraw, transfer and revoke — it has full authority over the account by definition. Non-custodial design moves that responsibility to you.

Use a hardware wallet for the account that holds meaningful size.
Verify the domain before every signature; phishing targets the wallet, not the protocol.
Treat seed-phrase requests as always hostile — no legitimate flow ever needs one.

What remains risky regardless

Perpetual futures are leveraged instruments. A mirrored basket can lose money, individual leaders can be liquidated, and execution differences mean your results will not match a leader's exactly. Isolation and caps shape the distribution of outcomes; they do not put a floor under it.

Past performance is not indicative of future results. Perpetual futures are leveraged instruments and carry a substantial risk of loss, including the loss of your entire position.

Methodology

Scoring and replacement are documented in full on How it works and in the Docs (Policy v3). In short: the Elite basket is sticky, emergencies remove a leader immediately, and soft issues accrue at most one strike per UTC day with three strike-days triggering replacement. Read how it works or the documentation for the full table.

Questions

Frequently asked

Could a hacked HyperMirror drain my account?

It could trade your account badly within the enforced caps. It could not withdraw or transfer funds, because that action is unavailable to an agent signature.

What is the single biggest risk in this model?

Market risk from leveraged positions, followed by your own wallet security. Custody risk is the one the architecture removes.

Does isolation help in a security incident?

Yes, structurally: sleeves are separate sub-accounts, so one leader's positions cannot net against or be funded by another's.

Keep reading

Agent vs API keysWho enforces the scope in each model.Approve and revoke safelyThe checklist that prevents most incidents.Risk disclosureWhat can go wrong, stated plainly.DocumentationScoring, allocation, isolation and replacement mechanics.

Start Autopilot View live performance

Diversified copy trading. On autopilot.

Score-weighted allocation across up to 10 elite Hyperliquid traders, each isolated in its own sub-account. Your funds never leave your account.