Kirkland Hints It Could Fine-Tune LLMs For Own Legal AI Model - Artificial Lawyer
Oh, marvellous. Kirkland & Ellis, the law firm so vast it reportedly bills like a small nation, has decided that the future of legal work involves fine-tuning open source LLMs on its own GPUs. A $500 million tech project, apparently, because nothing says 'cost-effective innovation' like dropping half a billion quid on servers and job postings for an 'AI Infrastructure Director' — a role that didn't exist five years ago and now sounds suspiciously like 'Head of Replacing Associates'. Big Law is indeed rethinking business as usual, as we noted last week when Freshfields started vetting software compliance across 100 countries. Now Kirkland wants in, but with on-premise GPUs, because the cloud is for plebs and the magic is in the silicon.
Let's look at those job postings, shall we? The AI Infrastructure Director will manage 'on-premise GPU environments' to fine-tune open source models. So the strategy is: take a model anyone can download, feed it a mountain of legal documents that the firm has presumably been billing clients to produce, and hope it doesn't hallucinate a motion to dismiss that cites the Yellow Pages. The $500 million figure is a hint that they're not messing about — or that they're paying the director a truly obscene salary. Either way, the partners will be just thrilled when the model starts second-guessing their billing multipliers.
What we're witnessing is the logical endpoint of a profession that has spent decades billable-hour-ing itself into a corner: the creation of a machine that can do the work for a fraction of the cost, while the humans who trained it nervously eye the exit. Kirkland's move isn't about efficiency — it's about insurance. A hedge against the day when clients realise that a well-tuned LLM can draft a contract in seconds, not weeks. The joke, of course, is that the model will be trained on past work product, which means it'll reproduce the same boilerplate, the same inflated clauses, the same everything — just faster. Progress, everyone. Cheers.