Tag: ChatGPT

Lawyers and AI

Lawyers And Artificial Intelligence: An Update

On By crowdfundattnyIn AI, Crowdfunding, PredictionsLeave a comment

I posted about lawyers and artificial intelligence in late 2023, predicting that AI tools would drive down the cost of legal services while making high-quality legal service available to more people. The great thing about predictions like that is that nobody can prove you were wrong until it’s too late. So far, however, it hasn’t happened.

I thought AI would enter the legal world through “intermediated” channels like Westlaw. With their enormous, curated databases of court cases, administrative rulings, and other source materials, as well as libraries of excellent legal forms drafted by top-notch lawyers, I expected companies like Westlaw to race quickly to the top, leaving “brute strength” tools like ChatGPT behind.

Since then, I’ve tried just about every AI tool on the market, including the most recent version of Westlaw’s AI tool, CoCounsel. Beginning each demonstration with high hopes, I am always left with great disappointment. 

Here are some things I’d expect from an AI tool for business lawyers:

  • Review Documents:  The tool should analyze an Asset Purchase Agreement or Operating Agreement and tell me (i) how it differs from “market” terms, and (ii) how it should be changed for the benefit of my client.
  • Summarize Documents:  The tool should summarize a legal document. One type of summary would tell me what’s in the document, at any level of detail I want. Another would prepare a summary I can use in an Offering Circular (e.g., “Summary of Management Agreement”).
  • Search Documents:  I’ve drafted approximately seven million Operating Agreements. If I’m looking for a clause I used two years ago, the tool should be able to find it.
  • Improve Documents:  The tool should review my document and point out ambiguities, inconsistencies, mistaken references, and logical gaps. 
  • Draft Sections of Documents:  If I’m drafting an IP License Agreement and need a section saying the Licensee is responsible for prosecuting infringement claims, the tool should produce one with a simple prompt.
  • Draft Whole Documents:  If I need a Rule 144 opinion, the tool should take me through the steps of preparing one, including the Certification from my client.
  • Legal Research:  The tool should vastly improve the process of legal research.
  • Reserve Flights:  Not necessary. 

In the earliest stages, I don’t expect an AI tool to produce great results. During a recent demonstration, the sales rep said, “You should view this as the work of a second-year lawyer.” Unfortunately, it was more like the work of a high school junior.

The good news is that brute strength tools like ChatGPT have improved dramatically. You still can’t rely on them – recently ChatGPT produced a quote from a court case speaking directly to my issue, but when I checked (always check), the quote was hallucinated – they are better than the intermediated tools, so far. 

When ChatGPT was released, many experts predicted that lawyers were the most vulnerable. Two and a half years later, that hasn’t happened, either. If you’re a lawyer, I guess that’s good news in a different way.

Questions? Let me know.

Artificial Intelligence

Anthropic: SPVs And The Investment Company Act

On By crowdfundattnyIn Investment Company Act, SPV10 Comments

I spend lots of time talking about special purpose vehicles (SPVs) and the Investment Company Act of 1940. Now we have a real-world example.

Anthropic was founded by Dario Amodei, who wrote the basic artificial intelligence model for OpenAI before leaving to start his own company. Once ChatGPT launched Anthropic has had no trouble raising money. They’ve raised $7.5 billion and counting in the last year.

In my humble opinion, the amount of money being thrown at Anthropic is insane. Most obviously, it demonstrates the psychological power of The Fear of Missing Out. More subtly, it represents the brokenness of venture capital culture. VCs have backed themselves into a position where they can no longer invest in businesses that are merely profitable. They need huge wins, grand slams. They bet a chunk of the farm on crypto/blockchain and lost. Now they need even bigger wins, or at least the promise of bigger wins, to keep their LPs writing checks.

Anyway, the flood of money created a problem for Anthropic that will sound familiar to many founders. The company was looking for billions, but many investors were able to invest “only” $30 – $50 million. The company didn’t want all those investors on its cap table.

So the company took the logical step:  it put the “small” investors in a separate company, an SPV, and admitted only the SPV to its cap table as a single investor.

Because its business is limited to holding securities in Anthropic, the SPV is an “investment company” under section 3(a) of the Investment Company Act. Yet it has not registered as an investment company. How does that work?

The answer is that it qualifies for the exemption under section 3(c)(1) of the Investment Company Act, section 3(c)(7) of the Investment Company Act, or both.

The exemption under section 3(c)(1) is available if the SPV has no more than 100 owners. That’s possible. If each owner invests $40 million you would raise $4 billion.

(NOTE:  the exemption under section 3(c)(1) allows 250 owners if the SPV follows a “venture capital strategy,” but this SPV was formed to invest in only one company, Anthropic.)

The exemption under section 3(c)(7) is available if each owner is a “qualified purchaser.” That term includes individuals with at least $5 million of investable assets, entities where all the individual owners have at least $5 million of investable assets, as well as other entities. I suspect the SPV qualifies under this exemption as well.

Thus, the SPV is an investment company under section 3(a), but is not required to register as such.

Finally, note that the discussion about the Investment Company Act doesn’t depend on how Anthropic raised money. It probably raised the money using Rule 506(b), taking the position that because everyone in that world knows everyone else, it had a “pre-existing relationship” with all its investors. But it could also have used Rule 506(c), assuming every investor is accredited. The point is that how you raise money and whether you need or qualify for an exemption under the Investment Company Act are unrelated.

I personally was not invited to invest in Anthropic. Imagine!

Questions? Let me know.

Artificial Intelligencer and crowdfunding

Artificial Intelligence And Crowdfunding Law

On By crowdfundattnyIn AI1 Comment

Like everyone else, I was shocked by the launch of ChatGPT. And like everyone else, I believed lawyers would be first on the chopping block. But I now have a far more optimistic view. I think AI will have a far more nuanced and ultimately beneficial effect for lawyers and their clients, including but not limited to Crowdfunding clients.

At first, I thought lawyers (or non-lawyers, gasp!) could type a question into ChatGPT and get a fully-formed legal product, whether a brief, a memo, a contract, or an obnoxious letter. But it turned out that neither ChatGPT nor its imitators is close to that, and I doubt they ever will be. 

Rather than crashing down the walls, AI is entering the legal profession through the front door. Lawyers are not interfacing with AI directly, by typing prompts into ChatGPT. Instead, the AI is being intermediated by existing legal resources. Through subscription, lawyers have access to extremely powerful online resources like the research tools at Westlaw and the high-quality legal forms at Practical Law. These services are themselves incorporating AI into their products, the same way Microsoft is incorporating AI into Office.

Today, for example, I can upload an Asset Purchase Agreement and get back all sorts of comments – what provisions are missing, as compared to a complete Asset Purchase Agreement, what provisions I should consider adding or deleting depending on whether I represent the buyer or seller, what’s “normal” for a given issue, correcting cross-references, letting me know which capitalized terms haven’t been defined, lots of other things. And I know that these comments and suggestions are coming, ultimately, from some of the best M&A lawyers in the country.

The AI is being intermediated by experts, who are using their experience and brains. In this way AI is not so much revolutionary as another step, if a large step, in the continuing evolution of legal resources.

Lawyers are using AI without even knowing it’s AI. And that’s perfectly normal. How many of us, flipping a light switch, think about electrons?

The result should be to make it easier for lawyers to produce a better product. Or to put it differently, to make high-quality legal work cheaper per hour.

I remember when lawyers thought email and fax machines would give them more leisure time and were shocked when they had the opposite effect. Email and fax machines allowed – actually, forced – lawyers to do more work. Rather than send a document overnight (itself an innovation) and wait for a response, the response was immediate.

The same will be true for AI, in my opinion. AI isn’t going to make lawyers redundant. Instead, with AI driving down the cost of a high-quality contract, existing clients will have more of their work done by lawyers, rather than trying to piece something together themselves, and people who have never used a lawyer will be able to afford one. The overall quality of legal work will rise, benefiting everyone.

Ever seen Apple’s original license agreement with Microsoft? The agreement was so awful, it basically allowed Microsoft to steal the GUI and launch Windows. And I am 100% sure the contract was so awful because a business guy at Apple wasn’t allowed to spend money on a lawyer.

You can see how this will translate to Crowdfunding. Upload a Form C and the AI will tell you what’s missing and suggest corrections and replacements. Upload the Y Combinator SAFE and the AI will tell you “Please read Mark Roderick’s blog post explaining why the Y Combinator form has to be changed for Reg CF.” Unfortunately, the Form Cs and other legal documents used for most Reg CF campaigns today are awful, like the Apple license agreement. I think AI will improve the quality of those documents and at the same time make Reg CF more accessible to more people. 

That’s a huge win.

Questions? Let me know.