Like all systems, smart contracts are only as good as the data they act on. A smart contract that receives bad or incorrect information from the network will still execute. On blockchain systems, this can be a huge issue as most transactions initiated by a human or a contract are irrevocable. Thus, if information is placed on a blockchain that is in error, fraudulent, or has some other deficiency, then a smart contract will still execute faithfully. Instead of expediting the proper functioning of the network, the smart contract would now be assisting in propagating an error.

To use the earlier example of shipping tires between FakeCar and Wheelmaster, what if during transit the boxes holding the tires were broken into and the tires replaced? If the worker at the FakeCar building scanned the boxes as received without checking each and every one, the smart contract would see this update and release escrow. The shipper would have their insurance bond returned, Wheelmaster would get paid, and FakeCar would still no longer have the wheels they ordered. To smart contract purists, this is how things should be. But in these cases, companies may instead refuse to use smart contracts or require additional layers of approval—essentially recreating the systems of old.

In designing smart contract systems, it is therefore critical that designers try and imagine every possible way things could go wrong. As with the DAO and other smart contract systems that have been used so far, small mistakes can have big consequences.

Many smart contracts involve some level of human interaction. For instance, multisignature wallets require multiple people to authorize a transaction before they will execute. These touchpoints introduce the same possibility for errors as old systems, but with the possibility of irrevocable consequences.