Enacted in 1976, the Hart-Scott-Rodino Act, or HSR Act, is an antitrust law that requires companies to file premerger notifications with the Federal Trade Commission and the U.S. Department of Justice before finalizing certain mergers, acquisitions or transfers of securities.
The FTC added new notification requirements in 2024, which took effect this February. These requirements require parties to submit substantially more information and documents than previously. These changes include more detailed information about ownership and potential competitive overlaps between parties, expanding the scope of potentially responsive documents in a filing.
The FTC estimates that the additional requirements could increase the filing time by up to 121 hours. Failure to comply with the new requirements could result in fines, with the maximum civil penalty for HSR violations increasing to $53,088 for each day of noncompliance.
These changes are expected to increase the time, effort and cost it takes to file premerger notifications, with the majority of the burden being placed on corporate legal departments that now have to review, analyze, and comb through more documents and data.
Generative AI is a potential solution for the new requirements as previous tools may not sort through the documents and data efficiently enough.
This process has gone from a very targeted inquiry in the past to trying to find a needle in a haystack, or rather trying to find relevant data in a larger set of documents, according to Erin Toomey, vice president, managing director and leader of the global investigations practice group at Epiq.
"It's going to take more time and cost more money to file these," Toomey told Law360 Pulse.
Epiq, which has used traditional forms of AI such as machine learning to help corporations with the HSR filing process in the past, said the number of companies it has helped in the previous year has nearly doubled in anticipation of the challenges with the new requirements.
"HSR deadlines have always been unforgiving," Fernando Delgado, head of the AI and analytics group for e-discovery and information governance provider Lighthouse, told Law360 Pulse.
Starting in 2018, Lighthouse used traditional AI methods to help customers quickly review data to comply with the old HSR regulations. This included a mixture of technology assisted review, a process known as TAR, which involves electronically classifying documents, and different models capable of processing data.
"HSR has been an area where we draw our innovation from because they are so exacting from a deadline standpoint," Delgado said. "You have to throw tech at that in a sensible way."
Under the new requirements of the HSR Act, Delgado said he noticed both the volume of data and the data types have increased in recent months.
Chris Cella, director of product management and e-discovery analytics at Sandline Global, told Law360 Pulse that the company is hearing concerns from legal teams about increasingly onerous challenges from the HSR requirements, including new deadlines and greater data collection demands. By expanding the scope of information and documents under the new HSR rules, Cella expects more e-discovery teams to get involved.
Several vendors have recently rolled out new offerings infused with generative AI, which is capable of creating original content, to help corporate teams keep up with the new HSR requirements.
In January, Epiq announced the general availability of the Epiq AI Discovery Assistant, a tool that uses generative AI models and chat features that allow companies to interrogate data. Toomey calls the tool a "review accelerator" that is needed because the HSR process can be nuanced and more complicated in some ways than normal litigation.
Epiq's AI Discovery Assistant ingests data and identifies events, entities, key parties and timelines, according to Toomey. Outside counsel and in-house teams can fine-tune the models to spot specific documents in each matter.
The company has already used the new tool in a few HSR requests and Toomey claims it has uncovered 80% to 90% of the relevant filing documents with very little oversight from the legal team.
"That empowers the legal team to be much more efficient and effective in their analysis," Toomey said. "Let the lawyers be lawyers, if you will."
Toomey adds that using generative AI for an HSR review can save time, as the company says it cut what would have been a three-week process for one company by one week. She also claims this could cut down on legal costs for HSR reviews.
To help clients deal with the new HSR Act regulations, Lighthouse has infused generative AI into its platform. Delgado said generative AI helps legal teams search for documents and determine which might have sensitive information that requires redaction.
Sandline will soon release a new AI-driven tool called ModernETL 2.0. Cella said the new tool handles what the company calls "loosely modern data," which encompasses data collection from different sources such as emails, collaboration platforms and text messaging. It also uses generative AI to summarize its findings.
Cella said that generative AI can help legal teams ensure accuracy given the new filing deadlines and potential high volume of data. He claims that generative AI won't replace human reviewers, but can help make sense of data in a way that previous analytics could not. This includes reducing the time it takes to conduct document reviews by narrowing the set of documents based on search criteria.
One concern with using generative AI during an HSR review is that the technology could be prone to producing inaccurate results, which is called a hallucination.
While providers cannot completely eliminate the chance of hallucination, Toomey said Epiq has mitigated those risks by doing more technology work in-house and by providing citations with answers.
According to Delgado, generative AI is typically accurate during an HSR review because it can find things that users deliberately ask for. Cella said that generative AI is usually accurate with basic tasks such as summarization. He also noted that human reviews are not impervious to making mistakes.
"Generative AI, if employed properly, can act as another reviewer," Cella said. "It can act as another set of eyes in a sense."
--Editing by Alex Hubbard.
For a reprint of this article, please contact reprints@law360.com.
