Over on the “Delaware Corporate & Commercial Litigation Blog”, Francis Pileggi recently flagged a concurring opinion by federal appellate judge Kevin Newsom in Snell v. United Specialty Ins. Co., (11th Cir.; 5/24). In that opinion, Judge Newsom explains how he used generative AI tools to help determine the “ordinary meaning” of terms at issue in a piece of litigation. In this particular case, the issue he sought to resolve was whether the installation of a trampoline involved “landscaping” within the meaning of a liability insurance policy.
After unsuccessfully trying to resolve this question by turning to the dictionary and photographs of the (very cool – see the pic on p. 7 of the opinion) in-ground trampoline itself, he decided to take a flyer on Chat GPT. After being intrigued by Chat GPT’s response to the general question “What is the ordinary meaning of ‘landscaping’?” the judge got more specific:
Suffice it to say, my interest was piqued. But I definitely didn’t want to fall into the trap of embracing ChatGPT’s definition just because it aligned with my priors. (Bad.) So, in what might have been a mistake—more on that later—we went ahead and asked it the ultimate question: “Is installing an in-ground trampoline ‘landscaping’”? ChatGPT responded as follows:
Yes, installing an in-ground trampoline can be considered a part of landscaping. Landscaping involves altering the visible features of an outdoor area for aesthetic or practical purposes, and adding an in-ground trampoline would modify the appearance and function of the space. It’s a deliberate change to the outdoor environment, often aimed at enhancing the overall landscape and usability of the area.
For good measure, I posed the same questions to Google’s Bard (since replaced by Gemini). The precise details aren’t particularly important, but the upshot is that both models’ answers indicated that the trampoline-related work Snell had performed—the excavation of the pit, the construction of the retaining wall, the installation of the mat, and the addition of the decorative wooden cap—just might be landscaping.
Judge Newsom then provides a thoughtful appraisal of some of the pros and cons of using AI large language models as a tool for resolving interpretive issues like this, and suggests that they may have a constructive role to play in judicial efforts to determine “the common, everyday meaning of the words and phrases used in legal texts.”
Francis wraps up his blog by summarizing the key takeaway from Judge Newsom’s opinion: “AI in the law is now mainstream when an appellate court includes references to how it was used as part of an analysis in connection with a published opinion.” Stay tuned.
The SEC’s Spring 2024 Reg Flex Agenda was released yesterday, and it looks like most of its rulemaking activity has been pushed out to after the election. Here’s where things stand on some of the potential SEC rules that we’ve been following:
As a reminder, these dates signify general timeframes. New final or proposed rules could come before or after the dates suggested in the agenda. The Reg Flex Agenda only gives insight into the priorities of the Chair as of the date it was submitted — it’s not a definitive guide for anyone trying to predict SEC rulemaking for purposes of specific board agendas, budgets and workflows.
In a recent issue of her “Deep Quarry” newsletter, Olga Usyvatsky noted that although comments on Schedule 13D and 13G filings have been relatively uncommon in the past, the Staff did recently issue a comment letter to Kimmeridge Energy Management LLC questioning the timing of a Schedule 13D filing. Kimmeridge responded to the comment and the Staff apparently didn’t have further comments on the issue, but Olga suggests that this may represent the start of a trend:
While SEC’s comments to Kimmeridge may look like a one-off, Corp Fin’s Director Erick Gerding highlighted compliance with the beneficial ownership rules as one of the priorities for the 2024 disclosure review program:
“Beneficial Ownership Reporting
On October 10, 2023, the Commission adopted amendments to modernize the rules governing beneficial ownership reporting.The Division staff is closely monitoring the implementation of these new rules. The Division staff will review selected beneficial ownership reports to assess compliance with the new, shortened filing deadlines and issue comments as necessary to improve required disclosures.”
Say it differently – we will likely see more SEC comments on 13D and 13G filings.
Jim McRitchie recently blogged about “Campaign GM” – a 1970 Ralph Nader-led shareholder proposal initiative and one of the earliest examples of shareholder ESG activism targeting a major corporation. The blog provides an overview of the campaign and discusses its long-term effects on governance at General Motors and its impact on the broader debate over corporate governance-related issues. It’s a really interesting introduction to a piece of corporate history that I knew very little about. Check it out if you get a chance.
Meredith blogged last week about the demise of the Chevron doctrine and some of its potential implications for the SEC. The conventional wisdom seems to be that the SCOTUS’s Loper decision is a “gut punch” to federal regulatory authority, and the decision has prompted wailing & gnashing of teeth on the Left and jubilation on the Right. However, a recent Law360 article by Dan Berkovitz says that the SCOTUS’s decision isn’t likely to have much of a practical impact on the SEC. Here’s the intro:
The demise of Chevron deference in Loper Bright Enterprises v. Raimondo will not make much of a difference for either the US Securities and Exchange Commission or the Commodity and Futures Trading Commission. The federal courts have a long history, both pre- and post- Chevron, of interpreting the federal securities and commodities laws without deference to the views of the SEC or the CFTC. These interpretations have ebbed and flowed between restrictive and expansive, depending on the philosophical composition of the courts.
In almost all major cases over at least the past two decades, especially in reviews of agency rulemaking, the courts have not granted Chevron deference to these agencies. Rather than charting a new approach, the court’s opinion in Loper reflects the current reality of judicial review for these agencies.
Dan Berkovitz isn’t alone in his assessment. This excerpt from a blog on the decision from the conservative think tank R Street Institute says that there was always less to Chevron deference than meets the eye:
There is also what you might call the dirty secret of judicial deference. Under Chevron, if a judge didn’t like an agency rule, they could usually just say that the rule was unreasonable (or that the statute wasn’t ambiguous) and achieve the same result as if there had been no deference. By contrast, even with the formal Chevron doctrine gone, it’s likely that judges will tend to implicitly defer to an agency’s interpretation more than they would if they had to come up with an interpretation of the rule from scratch.
The SEC’s recent enforcement action against RR Donnelley & Sons is the latest in a series of proceedings in which the agency has broadly interpreted the scope of the Exchange Act’s internal controls provisions. That approach has been sharply criticized by dissenting commissioners and by outside commenters, but in a recent “Radical Compliance” blog, Matt Kelly entertains the possibility that the SEC’s view of the world may be right.
Matt points out that Section 13(b)(2)(B) of the Exchange Act requires companies to maintain internal accounting controls “sufficient to provide reasonable assurances” that, among other things, access to assets is permitted only according to management authorization. He notes that in this enforcement proceeding, the SEC is taking a provision intended to apply to accounting fraud and applying it to cybersecurity – but as he explains in this excerpt, this isn’t necessarily an unreasonable position:
Is it really proper for the SEC to use its books-and-records provision in that manner? Honestly, I dunno. On one hand, we should remember that no actual fraud happened at Donnelley. No transactions were improperly recorded. The company didn’t even suffer a loss of data, since the data was only copied.
On the other hand, Donnelley was locked out of important IT systems. For example, some customers couldn’t receive documentation vital to vendor payments and disbursement checks. If this cyber attack happened in the real world, it would be akin to hooligans strolling into your building, changing the locks to the accounting department, and demanding millions if you want to get the set of new keys. A company that let something like that happen would certainly seem inept to most reasonable investors.
Critics of the SEC (and lord knows there are plenty around) would say the Donnelley case is a novel interpretation of anti-fraud rules, with the SEC basically nosing its way into cybersecurity regulation. That seems outside the SEC’s swim lane.
Then again, suppose those hackers had exploited sloppy cybersecurity controls to steal money from Donnelley rather than copying data, and then covered their tracks by altering the finance department’s banking records. (A frighteningly easy thing to do, by the way.) Few people would fault the SEC for raking Donnelley over the coals then. So why does this case feel a bit weird now, when money wasn’t stolen?
Matt suggests that we step back and look at the big picture – as technology has advanced, the controls required for strong financial reporting and those required for strong cybersecurity are converging into a single system focusing on access control. In this new reality, it’s essential to have strong controls to prevent unauthorized access to IT systems, rather than the historical norm of controls governing access to the accounting department and its physical books and records.
You might think that with all the negative attention from regulators about audit quality issues over the past few years, shareholders might be a little more hesitant to vote in favor of proposals to ratify auditors. However, according to a recent Ideagen/Audit Analytics report, if you thought that, you’d be wrong:
Throughout the last six years, our analysis on shareholder votes reveals that, on average, nearly 98% of total votes are cast in favor of auditor ratification. Shareholder votes filed between January 1, 2021 and December 31, 2023, continued that trend for a sixth consecutive year. Votes against auditor ratification comprised 1.7% of the total votes; abstained votes account for the remaining 0.4% of total shareholder votes cast.
In fairness, the percentage of proposals in which more than 5% of the outstanding shares voted against ratification of the auditors ticked up last year from 7% to 8%, but that still leaves 92% of proposals in which fewer than 5% of the outstanding shares were voted against ratification. Maybe shareholders ought to be a little more reluctant to toe the party line here, because a 2023 study found that higher than expected shareholder dissatisfaction with external auditors is associated with improved audit quality.
Yesterday, Corp Fin Director Erik Gerding issued a statement addressing concerns expressed by some registrants that the SEC’s rules requiring disclosure of material cybersecurity incidents in an Item 1.05 Form 8-K preclude registrants from sharing information beyond that disclosed in the 8-K with others, including contractual counterparties. Director Gerding’s statement clarifies that this is not the case, and that Regulation FD offers various alternatives for sharing this information without raising selective disclosure concerns:
There are several ways that a public company can privately share information regarding a material cybersecurity incident beyond what was disclosed in its Item 1.05 Form 8-K without implicating Regulation FD. For example, the information that is being privately shared about the incident may be immaterial, or the parties with whom the information is being shared may not be one of the types of persons covered by Regulation FD.
Further, even if the information being shared is material nonpublic information and the parties with whom the information is being shared are the types of persons covered by Regulation FD, an exclusion from the application of Regulation FD may apply. For example, if the information is being shared with a person who owes a duty of trust or confidence to the issuer (such as an attorney, investment banker, or accountant)or if the person with whom the information being shared expressly agrees to maintain the disclosed information in confidence (e.g., if they enter into a confidentiality agreement with the issuer), then public disclosure of that privately-shared information will not be required under Regulation FD.
The statement notes that while companies may be reluctant to share additional information about cybersecurity incidents with third parties, companies that follow the scope and requirements of the selective disclosure rules in Reg FD should not face undue impediments to mutually beneficial sharing of information regarding material cybersecurity incidents with third parties.
I recently saw a report quoting an OpenAI insider who estimates that there’s a 70% chance that artificial intelligence will destroy humanity. I guess that would worry me more if I didn’t put the odds of us doing that to ourselves without AI’s help at around 75% – and if the current iterations of AI didn’t have more in common with ’80s icon Max Headroom than with the HAL 9000 from “2001: A Space Odyssey.”
That being said, I’ve recently learned about one emerging use for AI that really does terrify me. Apparently, people are starting to use generative AI tools to prepare board minutes. A recent article in “The Boardroom Insider” flags this emerging practice, and this excerpt lays out some of the things that could go very wrong with relying on AI tools in this setting:
Potential downsides of this trend are apparent (and some are still to be realized). Recording of board meetings are always a legal bomb waiting to go off. The more it becomes a standard practice, the more likely someone will neglect to wipe all copies once minutes are finalized. While AI minuting apps note that their draft is only that — a draft for further human processing — what it retains and ignores can prove worrisome.
Further, once you get comfortable with letting AI do the minuting, you’re more likely to just send its digital take out for quick approval. AI “hallucinations” sneaking into the draft could be hard to spot. Finally, what if everyone on the board uses a recording to create their own AI summaries? This Tower of Babel approach could be a nightmare.
If you’re still willing to take the plunge, the article goes on to identify some AI tools that you might use to help generate board minutes. If you’re up for that, well, Godspeed! As for me, when it comes to the use of generative AI for board minutes, I’m firmly in Colonel Kurtz’s camp.
There’s some big Wu-Tang Clan-related crypto news that I’d like to share to close out the week. A few years ago, I blogged about how a “digital autonomous organization” or “DAO” named PleasrDAO had acquired the sole copy of the group’s legendary “Once Upon a Time in Shaolin” album that the feds grabbed from its original owner, fraudster Martin Shkreli.
PleasrDAO had big plans for the album, but those plans depended on its ability to persuade The RZA and Cilveringz to sign off on them. That sign-off was necessary because RZA opted to impose a unique restriction on any owner of Once Upon a Time in Shaolin when the album was announced in 2015 – whoever bought it would not be able to release it until 2103, 88 years following its release.
Well, it looks like PleasrDAO was successful, because according to this Bloomberg BusinessWeek article, it recently sponsored swanky listening sessions in NYC where attendees could hear selections from the album while “sipping artisanal cocktails.” If you missed the New York sessions, don’t despair – all you have to do is travel to New Zealand to catch the sessions being held at a Tasmanian museum through June 24th (in case you’re on the fence, they’re getting rave reviews).
So why is PleasrDAO holding these events? Well, this is where the tenuous connection to the federal securities laws that allows me to periodically blog about The Wu-Tang Clan comes into play. Here’s an excerpt from the article:
Perhaps not surprisingly, these exclusive sessions have coincided with the start of a campaign to wring more money from the album. On June 13, PleasrDAO started selling digital ownership stakes in Once Upon a Time in Shaolin for $1, entitling buyers to a short sampler from the album along with an encrypted file of the music that will remain locked—but maybe not until 2103, as originally promised. The collective says each sale will reduce the time it takes to make the entire album publicly available by 88 seconds. In short, the decades-long restriction is more fungible than most people might have assumed.
The article says that PleasrDAO raised $250K selling these NFTs in just four days, and cites a NY Times report as indicating that it would need to raise a total of $28 million to release the album to the public. It looks like PleasrDAO has attempted to structure this NFT to avoid having it classified as a security. Of course, that’s what all the NFT folks have said – and the SEC hasn’t always agreed. So, PleasrDAO would be wise to take some advice from the Wu-Tang Clan and “watch your step, kid, watch your step, kid, protect ya neck, kid!”
