Earlier this week, Mode.Inc, a Chicago-based consumer tech company that buys up mobile apps and pays users for their data, launched BizBuyData.com, a program pitched as a new revenue stream for law firms. The idea is that “[f]irms can license historical operational records to AI labs training workplace agents, earning from work their teams have already done without taking on debt or giving up equity.”
What counts as “operational records”? According to Mode, the good stuff is the “[i]nternal messages, emails and project histories” that show AI labs “how people work through problems and make decisions over time.”
So… the client file? Cool.
Let’s ask the obvious question: is this tech company asking firms to sell attorney-client privileged communications and attorney work product? Because the emails where a litigation team hashes out strategy, the Teams threads workshopping a client’s walk-away number, the project histories tracking how a matter got from intake to settlement… that is the work. The process Mode is so eager to capture is, in a law office, exactly the thing privilege exists to protect.
Mode’s answer is scrubbing. The company says it built tools that “remove sensitive information and replace names and other identifying details,” and that firms get to approve which systems and date ranges are included before anything is exported. Which, I mean… sure. But privilege doesn’t belong to the firm, it belongs to the client. Swapping a client’s name for a placeholder doesn’t make a strategy memo any less a confidential communication, and Model Rule 1.6 reaches all information relating to the representation, not just the parts with a name attached. The ABA’s Formal Opinion 512 already says lawyers generally need a client’s informed consent before feeding their information into self-learning AI tools, and that was about using AI, not selling the raw materials to train someone else’s.
And there is no expiration date on privilege! The Supreme Court made that clear in Swidler & Berlin v. United States — the Vince Foster notes case — holding that the privilege survives even the client’s death. A matter that closed in 2014 is every bit as privileged today and “historical” is not a loophole. Meanwhile, U.S. District Judge Jed Rakoff ruled earlier this year that a defendant who fed his lawyers’ advice into a consumer AI chatbot waived privilege over it, because sharing it with a third party is sharing it with a third party. Doing it for cash is not going to play better.
Mode co-founder and CEO Dan Novaes seems to see the lawyers coming:
“AI labs can’t scrape this kind of data. They have to license it from the firms that created it. But any company with a good legal team would walk away from a deal it can’t audit, and law firms should too. We put our own data through this process first, and we built it so firms can check the results before anything is shared.”
I’m not sure a firm can “audit” away its clients’ confidences.
Which brings us to the other question: what law firm is actually doing this? The eligibility bar is low, any firm with at least six employees using digital workplace tools, so the pitch reaches well past Biglaw, down to the small shops without a general counsel’s office or a risk committee to say no. And in case anyone needs a nudge, Mode is also paying referrers “the greater of $50,000 or 6% of a participating firm’s payment.” A $50,000 floor for making an introduction tells you how much Mode thinks this data is worth. It also means someone, somewhere, is going to be very motivated to pitch their managing partner.
But say a firm does take the time to filter all of it out — every privileged email, every piece of work product, every client confidence. What’s left? Calendar invites, conflict-check requests, and reminders that there’s leftover birthday cake in the 14th-floor kitchen. That isn’t particularly useful training data for an AI model that’s supposed to learn how professionals make decisions. The legal AI world has known for years that the bottleneck isn’t the tools, it’s the data, and there’s a reason it hasn’t been solved with a referral bonus.
Novaes is right about one thing: AI labs can’t scrape this data but that’s just because that’s how privilege works.
Kathryn Rubino is a Senior Editor at Above the Law, host of The Jabot podcast, and co-host of Thinking Like A Lawyer. AtL tipsters are the best, so please connect with her. Feel free to email her with any tips, questions, or comments and follow her on Twitter @Kathryn1 or Bluesky @Kathryn1
The post Tech Company Launches Program To Pay Law Firms For Their Emails, Privilege Be Damned appeared first on Above the Law.
Earlier this week, Mode.Inc, a Chicago-based consumer tech company that buys up mobile apps and pays users for their data, launched BizBuyData.com, a program pitched as a new revenue stream for law firms. The idea is that “[f]irms can license historical operational records to AI labs training workplace agents, earning from work their teams have already done without taking on debt or giving up equity.”
What counts as “operational records”? According to Mode, the good stuff is the “[i]nternal messages, emails and project histories” that show AI labs “how people work through problems and make decisions over time.”
So… the client file? Cool.
Let’s ask the obvious question: is this tech company asking firms to sell attorney-client privileged communications and attorney work product? Because the emails where a litigation team hashes out strategy, the Teams threads workshopping a client’s walk-away number, the project histories tracking how a matter got from intake to settlement… that is the work. The process Mode is so eager to capture is, in a law office, exactly the thing privilege exists to protect.
Mode’s answer is scrubbing. The company says it built tools that “remove sensitive information and replace names and other identifying details,” and that firms get to approve which systems and date ranges are included before anything is exported. Which, I mean… sure. But privilege doesn’t belong to the firm, it belongs to the client. Swapping a client’s name for a placeholder doesn’t make a strategy memo any less a confidential communication, and Model Rule 1.6 reaches all information relating to the representation, not just the parts with a name attached. The ABA’s Formal Opinion 512 already says lawyers generally need a client’s informed consent before feeding their information into self-learning AI tools, and that was about using AI, not selling the raw materials to train someone else’s.
And there is no expiration date on privilege! The Supreme Court made that clear in Swidler & Berlin v. United States — the Vince Foster notes case — holding that the privilege survives even the client’s death. A matter that closed in 2014 is every bit as privileged today and “historical” is not a loophole. Meanwhile, U.S. District Judge Jed Rakoff ruled earlier this year that a defendant who fed his lawyers’ advice into a consumer AI chatbot waived privilege over it, because sharing it with a third party is sharing it with a third party. Doing it for cash is not going to play better.
Mode co-founder and CEO Dan Novaes seems to see the lawyers coming:
“AI labs can’t scrape this kind of data. They have to license it from the firms that created it. But any company with a good legal team would walk away from a deal it can’t audit, and law firms should too. We put our own data through this process first, and we built it so firms can check the results before anything is shared.”
I’m not sure a firm can “audit” away its clients’ confidences.
Which brings us to the other question: what law firm is actually doing this? The eligibility bar is low, any firm with at least six employees using digital workplace tools, so the pitch reaches well past Biglaw, down to the small shops without a general counsel’s office or a risk committee to say no. And in case anyone needs a nudge, Mode is also paying referrers “the greater of $50,000 or 6% of a participating firm’s payment.” A $50,000 floor for making an introduction tells you how much Mode thinks this data is worth. It also means someone, somewhere, is going to be very motivated to pitch their managing partner.
But say a firm does take the time to filter all of it out — every privileged email, every piece of work product, every client confidence. What’s left? Calendar invites, conflict-check requests, and reminders that there’s leftover birthday cake in the 14th-floor kitchen. That isn’t particularly useful training data for an AI model that’s supposed to learn how professionals make decisions. The legal AI world has known for years that the bottleneck isn’t the tools, it’s the data, and there’s a reason it hasn’t been solved with a referral bonus.
Novaes is right about one thing: AI labs can’t scrape this data but that’s just because that’s how privilege works.
Kathryn Rubino is a Senior Editor at Above the Law, host of The Jabot podcast, and co-host of Thinking Like A Lawyer. AtL tipsters are the best, so please connect with her. Feel free to email her with any tips, questions, or comments and follow her on Twitter @Kathryn1 or Bluesky @Kathryn1
The post Tech Company Launches Program To Pay Law Firms For Their Emails, Privilege Be Damned appeared first on Above the Law.
