> For the complete documentation index, see [llms.txt](https://bitcoin-scaling-labs-docs.gitbook.io/ipc-btc-scaling-docs/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://bitcoin-scaling-labs-docs.gitbook.io/ipc-btc-scaling-docs/overview/trust-assumptions.md).

# Trust assumptions

#### Trust assumptions — clients

A client using a subnet places its trust in one place: that subnet's validator set. Concretely, the subnet's guarantees hold as long as an honest supermajority — two-thirds of the staked weight\
— is maintained. Under that condition the client gets both safety (any two honest clients see the same finalized order of transactions) and liveness (a submitted transaction is eventually finalized).

If corruption ever exceeds the subnet's threshold, the validators could stall or censor the client, reorder or rewrite the subnet's state, and put assets held there at risk. Even then, *checkpointing* bounds the damage. Because each checkpoint writes an immutable, ordered commitment of the subnet's state to Bitcoin, the honest history up to the last finalized checkpoint cannot be erased, and any later deviation from it is permanently anchored on L1 and publicly detectable.

A failure in one subnet does not extend to any other subnet. Each subnet keeps its own independent state and runs its own independent consensus, so a compromised subnet's misbehavior stays confined to it — it cannot alter the state or stall any other subnet. The only point of contact between subnets is cross-subnet transfers, and there the *firewall* property applies: a malicious subnet cannot send out more than it legitimately holds, because that limit is enforced by Bitcoin's own UTXO rules rather than by the subnet's validators. A client's root of trust therefore remains the validator set of the subnet it uses, sitting on top of Bitcoin L1 for settlement and finality.

***

#### Trust assumptions — validators

A validator's collateral is locked in a multisig spendable only by its own subnet's current committee, so it never depends on the validators of any other subnet — no outside committee can reach it, and a compromised subnet elsewhere cannot drain it. As the committee rotates, a validator trusts that the subnet remains safe — an honest two-thirds stake supermajority — at every committee across the subnet's lifetime.

A special case is a subnet whose validators are all operated by one entity: there, every committee throughout the lifecycle is that same entity, so the validator trusts no one but itself — a fully sovereign subnet — while still inheriting Bitcoin L1 for settlement.