Fed's Waller says AI at central bank has strict guardrails

Christopher Waller
Al Drago/Bloomberg
  • Key takeaway: Federal Reserve Gov. Christopher Waller said security measures designed to protect internal data can leave the Fed using slightly outdated technology, but are necessary for risk management. He added that the central bank uses AI for tasks including summarizing meetings.
  • Expert Quote: "In terms of research, you're going to have guardrails on it. We're probably not going to let you sit on a plane and run stuff on your laptop that has sensitive policy information on it. They won't even let me use it to clean up my email." — Federal Reserve Gov. Christopher Waller
  • What's at stake: Alongside its internal use of AI, the Fed has been closely monitoring how companies are deploying the technology for underwriting and compliance, as well as its implications for the labor market and monetary policy.

Federal Reserve Gov. Christopher Waller said Tuesday that guardrails are in place around how artificial intelligence is used in central bank research.

Processing Content

Speaking at a research-focused event hosted by the European Central Bank in Frankfurt, Germany, Waller said economists developing models at the Fed must follow protocols designed to protect sensitive data.

"There has to be guardrails on what happens, so we make sure that this stuff doesn't get out, or if we use inside information [that] it's not somehow used or helping companies use our inside information to develop their models," he said. "That's just a risk management problem for what we have to do."

Waller said the central bank operates a sandbox environment where economists can safely use AI models for research, though those tools may lag behind the latest technology because of security concerns.

"We have a bunch of different AI models that are provided to us from [Amazon Web Services] cloud, we have some guardrails," he said. "We may not have the most up to date, cutting edge model, because we had to check through and set things up."

He added that those guardrails may leave researchers "a little behind doing work at the bank," but said the tradeoff is necessary "to protect against if other things happen."

Waller described AI as an "amazing tool" and pointed to use cases at the Fed, including summarizing meetings.

"If you want to do an executive summary of the whole meeting, run the transcript through it. It gives you an executive summary, a very fast and efficient way to do the minutes of our meetings," he said. "There's lots of really nice ways you could use this stuff in the central bank to make us all more efficient."

Still, Waller emphasized the importance of limiting how and where AI tools are used, particularly when sensitive internal information is involved.

"In terms of research, you're going to have guardrails on it," he said. "We're probably not going to let you sit on a plane and run stuff on your laptop that has sensitive policy information on it.

"They won't even let me use it to clean up my email," he quipped.

Alongside its internal use of AI, the Fed has been closely monitoring how companies are deploying the technology and what effects it could have on the labor market.

As AI tools rapidly proliferate, companies have rushed to integrate the technology in an effort to streamline operations and reduce costs.

Several Fed officials, including Governors Lisa Cook and Michael Barr, have warned that AI could disrupt parts of the workforce, though they say it may be too early to determine the full impact.

Barr said rapid implementation of AI tools by employers could make some jobs obsolete.

"You may be beginning to see a little bit of that in employment data — not a lot, but a little bit of it in some sectors," Barr said during an appearance late last year. "It might be an area to watch in the future."


For reprint and licensing requests for this article, click here.
Federal Reserve Artificial intelligence Economy
MORE FROM AMERICAN BANKER
Load More