When AI Wants a Cut: What OpenAI’s Pharma Move Means for Legal Tech

OpenAI’s recent talk about taking a share of drug revenues if its models help with discovery might sound like something only pharma (and similar big R&D industries) needs to care about. It isn’t, it’s a glimpse of what happens when AI stops being a tool and starts being treated as a partner. For legal tech, that idea isn't too far off really.

From tool to partner

In pharma, the logic’s simple. If an AI system helps discover a drug, that’s not support work, that’s creation. The upside’s huge, so OpenAI’s exploring deals that look more like partnerships than software licences.

Now let's have that logic applied to law.

• If an AI reviewer halves diligence time, is that efficiency or co-production?
• If an AI drafter builds half a document, does that count as support or authorship?

It sounds far-fetched, but once you can measure what AI contributed, someone will start arguing over who owns the value.

Legal's weird middle ground

Legal's always defined value through two things: time and expertise. The billable hour put a price on effort, and professional qualification gave that effort credibility. For years, that pairing has underpinned how firms justify value, fees, and reputation.

AI challenges both.

• Time stops being a fair measure when software can read thousands of pages, flag risks, and draft summaries before anyone’s finished their first coffee.
• Expertise blurs once models start reflecting a firm’s own reasoning patterns, by learning tone, risk preferences and drafting habits that used to be the mark of seniority.
• Reputation shifts too, as clients begin judging output on speed, consistency, and accuracy rather than on name alone.

That’s where the OpenAI idea starts to hit. The legal business model was built on human effort, not shared creation. Once AI starts contributing directly to the work product, the old boundaries, between creator and tool, between professional judgement and system output then start to fade.

The profession’s rules were never designed for this kind of partnership, and they’ll need to catch up fast.

Why it could happen in legal

The main things are already in place:

1. You can measure contribution
AI systems in law are benchmarked every day. How many clauses extracted, how many documents summarised, how many hours saved. Once you can measure, you can argue about credit.

2. Data’s the fuel
Firms feed proprietary data into vendor models, making them smarter. Vendors will say that data adds lasting value, so they should share in the outcome.

3. The economics are changing
Compute’s expensive and open models are closing the performance gap. Vendors will look beyond licences to keep margins healthy. Performance-based fees or revenue shares are a logical next step.

The domino problem

Add to that the current market. Legal tech is now full of heavily funded tools under pressure to prove returns. If one vendor makes a move toward revenue-sharing or profit-linked pricing and it works, others will copy.

By then, saying "we’ll just switch to something else" won’t be realistic, because these tools are no longer sitting behind the scenes, they are the workspace. Teams draft, review, and collaborate inside them. Replacing one means retraining users, rebuilding automations, and losing months of embedded habits.

That’s why firms need to think now about how tightly they let a single tool wrap around their processes. AI integration should be seamless but swappable. Use shared data layers, keep your own context stores, and avoid making the vendor’s logic your operating system.

It’s not about resisting innovation, it’s about keeping your options open when pricing models start to shift.

A real-world parallel

Let's say you have a team performing due diligence using an AI tool that reads 20,000 contracts and flags 98 percent of change of control clauses correctly. What took two weeks now takes two days.

Next renewal, the vendor suggests a new deal: smaller base fee, plus a three percent share of the cost saving compared to last year. It’s attractive on paper.

But...

• Who proves the saving?
• What if the model misses a clause that costs the client later?
• Would sharing the saving breach SRA rules?
• Does the vendor’s financial interest compromise independence?

That’s all before you get to client confidentiality. To prove their share, the vendor needs data, now that could expose privileged material.

If you’re locked into their platform as your main workspace, walking away isn’t simple.

The legal and regulatory wall

Under SRA rules, non-lawyers can’t share in legal fees unless the firm’s an Alternative Business Structure. Even then, independence is a red line. A vendor with money riding on the outcome could influence strategy, even indirectly.

Then there’s liability. If a vendor wants a slice of success, are they also taking a slice of risk? In pharma that’s managed with milestone payments and insurance. Law doesn’t have that setup.

The attribution problem

To even talk about value sharing, firms would need proper attribution. That means:

• Fixed evaluation sets for each workflow.
• Logs showing which model, prompt, and data set were used.
• Evidence of what the model contributed to the final output.
• Change control when the model updates.

Without this, any talk of performance fees is smoke and mirrors. With it, at least there’s a record of what really happened.

What firms and vendors should be doing

Firms should be getting ahead of this now, not waiting for the first contract renewal where it becomes an argument. Set your stance clearly and build your governance around it.

Firms:

• Define what contribution means and how it’s proved.
• Protect independence in writing.
• Start building attribution capability now.
• Keep your integrations modular so switching vendors doesn’t cripple operations.

Vendors:

• Be clear about what data you keep and how it’s used.
• Avoid ownership claims that scare buyers.
• Prove impact with evidence, not marketing.
• If you want upside, accept shared accountability when things go wrong.

Why it matters now

Pharma’s deals show how quickly business models shift once AI’s value is measurable. Legal tech’s not far behind. Managed services already use outcome-based pricing. It’s only a matter of time before someone tests a performance-based model with AI.

When that happens, dependency on a single vendor will make those negotiations messy. Firms that built flexibility in early will hold the leverage. Those that didn’t will have to accept the terms they’re given.

The next big test for AI in law won’t be accuracy or creativity. It’ll be economics.
Once a machine can prove it contributed, the question becomes who owns that contribution and whether the firm still controls the deal.

Pharma might be the first industry where AI takes a seat at the profit table. Law won’t be far behind. The firms that keep their tech replaceable will be the ones still running the table when that shift arrives.