Really interesting read, especially a year on as Harvey has now doubled its valuation to $11bn and the competitive dynamics you flagged are very much still in play. Anthropic's partnership with Freshfields is now another variation of it.
One thing I find interesting now is that Harvey seems to be aware of this which is why the BigLaw Bench and their ELO evaluation work feels more than just marketing to me. I read them as Harvey trying to make the software layer defensible on its own terms, rather than just relying on whatever the foundation models can do generically.
Great analysis, Zach. Harvey's $75M ARR success story has a compliance blind spot most firms haven't considered.
Every Harvey query creates discoverable logs at OpenAI that survive litigation holds. The NYT v. OpenAI case proved that provider retention policies don't protect client data when courts freeze deletion. Marketing claims of "we delete your data" become meaningless under preservation orders.
Harvey's API wrapper architecture means client communications flow directly through OpenAI's systems. At $75M ARR scale, that's massive volumes of privileged content sitting in discoverable third-party logs. The bigger Harvey gets, the bigger the target for subpoenas.
Harvey proves legal AI demand is real. But it also proves the need to solve the retention risk before the first major discovery motion hits.
I have been trying out a few of these legal AI tools recently. What stands out is how quickly they have gone from clunky assistants to something you could actually rely on for first drafts.
Harvey knows that even the failure mode here is an acquisition offer by OpenAI (it's seed investor). Maybe not Jon Ive money and glamour shots with Sam Altman, but at least enough to make investors whole. So if there’s a chance at hyperscale for a vertical app on law, Harvey AI has a shot at it. I’m rooting for them—not just because we share a name, but because the two large publishers have boxed this area in for decades and it’s nice to have a little competition.
But bigger picture - If you take this a step even further, I think entry level jobs and associate attorneys are going to have an existential crisis moving forward.
“We have to will it to be, because otherwise we’re going to have deep societal upheavals that I think many in our elite are just really ignoring,” Karp said.
- #1 In a recent interview, Alex Karp, CEO of Palantir, warns urgently of the social upheaval that the rapid use of artificial intelligence could bring. His central message: if AI is introduced unregulated and without political design, there is a risk of profound social destabilization - above all through the loss of traditional entry-level jobs.
- #2 The jobs that are particularly at risk are precisely those that people have previously used to gain access to economic participation, for example through hard work, training or initial work experience. Karp sees this development not only as an economic problem, but also as a social one: if broad sections of the population feel that there is no longer an “entry-level job”, this undermines trust in the democratic system in the long term.
- #3 He sharply criticizes the fact that many elites - in politics, business and technology - do not recognize the true extent of this impending development. While they talk about the efficiency gains and innovative power of AI, Karp says they are ignoring how quickly entire professions could become redundant. The change will not happen in decades, but in a few years. This makes it all the more urgent to act now.
- #4 Karp is not fundamentally against AI. On the contrary: he certainly sees positive potential - for example in increasing productivity and reducing the burden of routine tasks. However, these benefits can only be realized if society is prepared to actively adapt to the change. This means investing in further training, a fair redistribution of technological gains and a political vision of what the new world of work should look like in concrete terms.
- #5 Otherwise, the gap between the winners and losers of digitalization will become so wide that it could tear apart the social fabric. His warning is therefore also an appeal: those who fail to shape the future now risk a future in which technology does not liberate, but divides.
I'd be interested to see which firms Palantir has on retainer. When Palantir gets rid of its in-house legal department and cuts ties with white shoe firms, I'll believe it more. I believe we could see real disruption, but I like that because it could mean new opportunities for new entrants. This may be a bit fence-sitty but I think AI will bring both boom and doom. I'm looking for the boom and trying to avoid the doom.
Really interesting read, especially a year on as Harvey has now doubled its valuation to $11bn and the competitive dynamics you flagged are very much still in play. Anthropic's partnership with Freshfields is now another variation of it.
One thing I find interesting now is that Harvey seems to be aware of this which is why the BigLaw Bench and their ELO evaluation work feels more than just marketing to me. I read them as Harvey trying to make the software layer defensible on its own terms, rather than just relying on whatever the foundation models can do generically.
Whether that holds is the big question.
If you think Freshfields is impressive, take a look at the Kirkland announcement from yesterday.
Great analysis, Zach. Harvey's $75M ARR success story has a compliance blind spot most firms haven't considered.
Every Harvey query creates discoverable logs at OpenAI that survive litigation holds. The NYT v. OpenAI case proved that provider retention policies don't protect client data when courts freeze deletion. Marketing claims of "we delete your data" become meaningless under preservation orders.
Harvey's API wrapper architecture means client communications flow directly through OpenAI's systems. At $75M ARR scale, that's massive volumes of privileged content sitting in discoverable third-party logs. The bigger Harvey gets, the bigger the target for subpoenas.
Harvey proves legal AI demand is real. But it also proves the need to solve the retention risk before the first major discovery motion hits.
I have been trying out a few of these legal AI tools recently. What stands out is how quickly they have gone from clunky assistants to something you could actually rely on for first drafts.
Harvey knows that even the failure mode here is an acquisition offer by OpenAI (it's seed investor). Maybe not Jon Ive money and glamour shots with Sam Altman, but at least enough to make investors whole. So if there’s a chance at hyperscale for a vertical app on law, Harvey AI has a shot at it. I’m rooting for them—not just because we share a name, but because the two large publishers have boxed this area in for decades and it’s nice to have a little competition.
But bigger picture - If you take this a step even further, I think entry level jobs and associate attorneys are going to have an existential crisis moving forward.
“We have to will it to be, because otherwise we’re going to have deep societal upheavals that I think many in our elite are just really ignoring,” Karp said.
Here is why: https://x.com/kimmonismus/status/1931730962586685806
- #1 In a recent interview, Alex Karp, CEO of Palantir, warns urgently of the social upheaval that the rapid use of artificial intelligence could bring. His central message: if AI is introduced unregulated and without political design, there is a risk of profound social destabilization - above all through the loss of traditional entry-level jobs.
- #2 The jobs that are particularly at risk are precisely those that people have previously used to gain access to economic participation, for example through hard work, training or initial work experience. Karp sees this development not only as an economic problem, but also as a social one: if broad sections of the population feel that there is no longer an “entry-level job”, this undermines trust in the democratic system in the long term.
- #3 He sharply criticizes the fact that many elites - in politics, business and technology - do not recognize the true extent of this impending development. While they talk about the efficiency gains and innovative power of AI, Karp says they are ignoring how quickly entire professions could become redundant. The change will not happen in decades, but in a few years. This makes it all the more urgent to act now.
- #4 Karp is not fundamentally against AI. On the contrary: he certainly sees positive potential - for example in increasing productivity and reducing the burden of routine tasks. However, these benefits can only be realized if society is prepared to actively adapt to the change. This means investing in further training, a fair redistribution of technological gains and a political vision of what the new world of work should look like in concrete terms.
- #5 Otherwise, the gap between the winners and losers of digitalization will become so wide that it could tear apart the social fabric. His warning is therefore also an appeal: those who fail to shape the future now risk a future in which technology does not liberate, but divides.
I'd be interested to see which firms Palantir has on retainer. When Palantir gets rid of its in-house legal department and cuts ties with white shoe firms, I'll believe it more. I believe we could see real disruption, but I like that because it could mean new opportunities for new entrants. This may be a bit fence-sitty but I think AI will bring both boom and doom. I'm looking for the boom and trying to avoid the doom.