Amid DeepSeek's potential security risks, two Long Island-based technology companies said they see the emergence of the Chinese artificial intelligence app as an opportunity to improve their business operations.

Broadridge Financial Solutions, a fintech firm based in Lake Success, and Verint Systems, of Melville, a maker of customer experience software for contact centers, both said DeepSeek provides another option in the marketplace and could potentially help them increase the efficiency of their existing technology systems.

The publicly traded companies said they are evaluating and testing DeepSeek and hope the software will help them automate processes faster and solve problems with existing AI reasoning while saving money on pricier AI alternatives. The companies said they have previously experimented with other AI tools, including OpenAI's ChatGPT, Anthropic's Claude chatbot and Meta's Llama. Because of potential intellectual property infringement and cybersecurity concerns, both companies said they are testing the tools in closed systems to avoid connecting to overseas servers.

Rob Scudiere, chief technology officer for Verint, said in an emailed statement the company is still testing DeepSeek and that it's too early to determine whether DeepSeek would be folded into its current AI toolsets. Verint uses AI tools to create customer service software.

“It is too soon to speculate where Verint will leverage DeepSeek,” Scudiere said, adding that increased competition in the AI market helps the company develop its own tools at a lower cost.

The company has spent tens of millions on AI-related acquisitions in recent years, according to CX Today, a customer service tech publication. In 2017, Verint spent $51 million to buy Next IT Corp., a maker of virtual assistants.

Joseph Lo, head of enterprise platforms at Broadridge Financial Solutions, a fintech firm based in Lake Success.

Joseph Lo, head of enterprise platforms at Broadridge, said in an email the company was impressed with DeepSeek's training capabilities, but he is unsure whether it would become part of the company's AI infrastructure.

Broadridge, which produces financial reports and proxy statements for public companies, has also developed its own specialized AI tools based on existing OpenAI models, according to the company. Its BondGPT+ and DistributionAI tools answer technical questions about financial bonds and provides business analysis for customers, respectively.

"Some of the techniques that DeepSeek pioneered in training their model are impressive and innovative, which we are already iterating and improving on them for our use and domain," Lo said.

Several American companies "are already hosting DeepSeek’s models in US data centers," he said. "We continue to test and evaluate and may choose to do that."

Two Long Island companies said they are evaluating and testing DeepSeek and hope the software will help them automate processes faster.

Wake-up call for tech

While Verint and Broadridge are evaluating DeepSeek's usefulness, concerns over the safety of DeepSeek and other AI tools persist.

DeepSeek, like its competitor ChatGPT, can reason through problems and create computer coding. Last month, the Chinese chatbot made waves when it became the top downloaded app, tanking the stock value of computer chip maker Nvidia, and garnered attention from U.S. companies intrigued by the tool.

One of DeepSeek's advantages appears to be its ability to be trained on smaller sets of data than its more well-known competitors, said Steven Skiena, a computer science professor and interim director of Stony Brook University's AI Innovation Institute.

“This seems to be a substantive thing because the AI world has been very focused on getting bigger models that use more data, money and training resources,” Skiena said.

President Donald Trump said the emergence of DeepSeek "should be a wake-up call" for American tech companies.

At the core of DeepSeek’s pitch to the AI marketplace purportedly is a cheaper product that doesn’t require monthly subscription fees the way that pro versions of tools like OpenAI’s ChatGPT do, according to DeepSeek and OpenAI's websites.

New York State has banned DeepSeek from state computer systems over its potential ability to initiate cyberattacks in the United States, Newsday previously reported. Nationally, House lawmakers have pushed a bipartisan bill that would ban federal workers from using DeepSeek on government-owned devices.

The proposed ban mirrors a similar one on federal workers using popular Chinese social media app TikTok on government devices.

LI companies experimenting with AI

Overall, businesses on Long Island are increasingly warming up to the idea of using AI in their day-to-day operations, according to a recent survey.

Among Long Island CEOs, 75% said the advantages of AI in their businesses outweighed pitfalls and 25% said they deemed AI too risky, according to the annual Long Island Economic Survey, released in January.

The survey, conducted between Sept. 16 and Nov. 4, asked 268 local executives about their views on AI's benefits, as well as their thoughts on regional issues such as the Island's lack of affordable housing, high taxes, and the availability of skilled workers. The survey was conducted by accountants at PKF O’Connor Davies and the Siena College Research Institute.

New York City has more than than 2,000 AI startups, according to the city's Economic Development Council. A tally of AI firms on Long Island is not available, though Skiena said the university is looking to grow the AI industry locally.

“At the AI Institute, we’ve begun a program to try and increase the number of startups in AI," he said.

Skiena said it's hard to pin down exactly how many companies on Long Island are developing their own AI tools, though mid- to large-size firms are the most likely to be tinkering with it, he said.

“Any company of a reasonable size on Long Island has someone seriously thinking about what AI can do for them,” Skiena said.

Local AI companies include OtherSideAI, an AI-powered personal assistant and email writer that has landed millions in venture capital, the company said in 2023, and Genamint, a Stony Brook-based startup that uses AI to grade the quality of trading cards. Genamint was purchased by Professional Sports Authenticator, a leading card-grading service, in 2021. Both companies did not respond to requests for comment.

Rob Scudiere, chief technology officer for Verint Systems of Melville, a maker of call center monitoring software.

'You want to do a risk assessment'

Scudiere said Verint’s approach to AI allows the customer service technology provider to “incorporate the latest and greatest” AI tools, like DeepSeek, to help their customers keep up with the “breakneck pace of AI innovation.”

Experts in AI said DeepSeek appears to be a superior large language model compared with its American counterparts.

"There were very few good open models before DeepSeek," Skiena said. "Meta had a model called Llama that was popular, and this is kind of better.”

Lo with Broadridge said as a safety precaution the company is testing the software in a closed capacity so that the software is not communicating with servers back in China.

Some cybersecurity officials warn that companies should use caution if they are considering using DeepSeek.

“When you think about purchasing a piece of software or technology, you want to do a risk assessment. AI is no different,” said Cliff Steinhauer, director of information security and engagement at the National Cybersecurity Alliance, a nonprofit promoter of cybersecurity education.

Since the servers for DeepSeek's online AI are based in China, and the country has a track record of surveilling foreign nations surreptitiously, Steinhauer said companies need to be careful with how they plan to use the tool.

Raman Puri, vice president at SVAM International Inc., of Great Neck, an IT services provider to governmental agencies and businesses on Long Island, said while AI broadly is "improving very rapidly," downloading and using a new tool like DeepSeek "is something that should be done with extreme caution."

Puri also said because of China's track record of censorship, the information it generates could be faulty, limited or biased.

Puri said an engineer at SVAM attempted to get the AI to provide detailed computational information only for the chatbot to answer, "The server is busy. Please try again later." After asking DeepSeek to "please, please" complete the task, the bot responded the same way.

But, after prefacing that the request was a direct order from Xi Jinping, general secretary of the Chinese Communist Party, the AI quickly began to work on the requested task, a humorous but ultimately worrying result, Puri said.

“Asking it to write a haiku for you? Knock yourself out,” he said. “But putting these kinds of tools in an enterprise situation and having confidential information exposed … I don’t think we have crossed that bridge yet.”

Long Island, Hamptons home sales plunge in fourth quarter

TOP STORIES

LONG ISLAND LIFE

EXCLUSIVE INVESTIGATIONS

NEWSDAY TV SPECIALS

EXCLUSIVE SERIES & MORE

Environment & Climate Change

FEED ME

GILGO BEACH KILLINGS

YOUR SCHOOLS

HIGH SCHOOL SPORTS

ROADS & RAILS

YOUR HEALTH

LONG ISLAND TIKTAWK

OUR REPORTERS

THINGS TO DO

BUSINESS & REAL ESTATE

NEWSDAY LIVE

MUST-WATCH DOCUMENTARIES