ZURICH. For decades, the pursuit of lost, stolen or concealed financial assets was labour-intensive, and often uneven. Illicit actors moved faster than the people chasing them, from layered offshore companies to cryptographic obfuscation. Manual ledger reviews and long cross-border disputes left even skilled legal teams at a dead end. Financial restitution is now changing, and the change is being driven by artificial intelligence.
The modern financial system is fragmented. Capital moves across global networks in milliseconds. When funds disappear into shell entities or onto a decentralised blockchain, the volume of data is beyond what a person can read by hand. Machine learning changes that arithmetic. Models can take in very large datasets, map transaction patterns that would not show up in a sample, and follow the digital traces of illicit movements across thousands of international accounts at the same time.
Where the algorithm meets the brief
Technology, on its own, does not win a case. The practical advance is the joining of data science to legal strategy. Many institutions are still discussing what these tools might do in theory. A smaller number are already using them in live files, and some of that work is starting to shape how courts are asked to act.
One example is the practice of Lucas Olsen and Oscar Keller at SwissLawmakers. They have not treated artificial intelligence as a forensic add-on. Predictive analytics and deep-learning models sit inside the legal framework of the file. The point of the method is the gap it tries to close: raw algorithmic output on one side, and something a court can be asked to order on the other.
From a flag to a freezing order
In that method, the model does the first job. It flags anomalies. It untangles funds that have been routed through cryptocurrency mixers or offshore havens. The next job is translation. Keller and Olsen use those findings to build the legal structure of the case: algorithmic results turned into evidence, then used to seek international freezing injunctions, to pierce complex corporate veils, and to coordinate litigation in more than one jurisdiction.
That last step is the one exchanges and banks already understand. A civil injunction, paired with a forensic link to a wallet, is what can lock stolen Bitcoin, ETH or USDT before an off-ramp pays out. A chart of hops, however complete, is not an order. The counsel who can carry a model’s output into a prayer for relief are the ones who turn a trace into a hold.
Parity, not a replacement
That combination is where this part of financial law is heading. A reactive investigation, which starts after the money has already moved and then tries to catch up by hand, is a poor match for a market that settles in milliseconds. A strategy that starts with the data, and only then drafts the relief, is the one practitioners such as Olsen and Keller are putting to work.
Artificial intelligence is not taking the lawyer’s place. It is equipping the counsel who use it with a way to restore some parity, so that the speed of digital finance does not automatically outrun the reach of a court.