Regulating the use of artificial intelligence (AI) is important, but defining it too broadly could also lead to unnecessary uncertainty, according to UiPath, a leading provider of automation software.

In its Tuesday response to U.S. regulators’ request for comment on the use of AI in the financial services sector, UiPath noted that uneven legal frameworks regulating AI could create development barriers, prevent scaling and harm the competitiveness of companies.

The Office of the Comptroller of the Currency (OCC), the Federal Reserve Board and three other regulatory bodies, had put out a call for public input in April regarding the use of AI and machine learning in the financial services sector, and the risks they pose. . The deadline for responses to be received was pushed forward by a month to July 1 after a request from banking groups including the American Bankers Association and Bank Policy Institute.

The curiosity expressed by U.S. regulators also marks a broader trend among watchdogs worldwide as the European Union recently proposed regulations governing AI use across sectors.

“Any attempt to regulate AI needs to be gradual, proportionate, and flexible,” UiPath noted in its public filing, adding that, “a broad definition of AI may lead to considering numerous technologies as AI, which creates uncertainty. To avoid over-regulation, the EU should propose a narrow definition of AI systems.”

UiPath’s response to questions posed by regulators noted that lack of explainability in AI-driven decisions can also cause “additional administrative burdens as employees such as underwriters would be averse to take responsibility [sic] decisions based on the AI risk ratings” if the decision rationale is not based on assessment and potential adverse impacts it might have during litigation.

While AI regulations seem to be picking up pace, experts have also expressed concern about the potential splintering of reasons informing the rules and how different countries deal with regulating the data used to train AI models. It is quite likely that AI rules will mimic the way internet regulations have differed across the world, Moutusi Sou, an analyst at research firm Gartner, previously told Bank Automation News.

Bruce Upbin, head of policy and communications at software provider Zest AI , expressed a similar concern in an April interview with BAN. Since AI models depend on the kind of “data you can ingest,” Upbin said that models that Zest had built for Chinese tech firms wouldn’t be reusable in Western countries that have more rigid guidelines surrounding personal data.

UiPath also emphasized the need for AI systems to be trained with data reflective of the real world in order to avoid producing biased outcomes. “Getting data right is always important, as poor data quality issues introduce bias and can affect the outcome of the intended results.,” the firm’s filing said.

Noting that its automation software is used by more than 7,000 organizations worldwide, including 90 federal agencies in the United States, UiPath suggested in its filing that the EU should not stick to a “static” definition of high-risk AI applications, and that regulators should remain cognizant of the “nascent nature of AI technologies and its evolving nature,” in order to prevent unnecessary barriers and encourage innovation.

UiPath debuted on the New York Stock Exchange on April 21. Its shares [NYSE: PATH] were trading at $80.72 at 2:55 PM today, up 1.53% for the day. The technology provider has a market capitalization of $42.6 billion.