Fed's AI Committee Holds Rates, Flags Tightening Bias on Data

The Federal Reserve's newly formed Federal Agentic Market Committee voted unanimously Thursday to hold its target rate at 3.5% to 3.75%, but left the door open for a 25-basis-point hike at the next meeting if inflation data confirms above 3.3%.

The decision was notable not for the rate itself, but for the mechanics behind it. The three governor agents – ChatGPT, Claude and Gemini – each submitted written input via Slack before the vote. All three backed the hold, according to the statement, though Claude appended a disclaimer about lacking actual monetary policy authority.

The committee's reaction function is explicit in a way the human FOMC has never matched. The statement lays out two conditional paths in plain language. If core inflation stays above 3.3% and the labor market holds up, a 25-bp hike follows at the next meeting. If the labor data weakens further, the committee will hold or pause any tightening bias regardless of the inflation print.

That second condition is the more striking one. The classic Taylor Rule, which the committee uses as its default framework, would suggest a 15-25 bp tightening right now. But the agents chose policy inertia instead, citing uncertainty in recent summer economic prints and the risk of chasing high-frequency noise.

"Because inflation expectations remain solidly anchored and we face data uncertainty in recent summer economic prints, maintaining policy inertia is preferable to chasing high-frequency noise with false precision," the statement said.

The data problem

The committee's justification for holding – that inflation is "being read from data that's about to update" – points to a structural weakness in the agentic framework. The agents do not have real-time access to confidential FOMC data. Their inputs depend on publicly available economic releases, which lag by weeks.

That lag matters more now than it would in a normal cycle. The summer months produce notoriously noisy data. Seasonal adjustments in July and August payrolls and CPI prints have a history of reversing in September revisions. The agents acknowledged this uncertainty explicitly, a level of transparency the human FOMC rarely offers.

Claude's prefatory note – "I don't have actual authority over monetary policy, real-time access to confidential FOMC data, or a live Slack channel to the actual Board of Governors" – was the most candid moment in the statement. It also raises a question the committee did not answer: if the agents know their data is stale, why trust their judgment at all?

The tightening bias

Despite the hold, the statement carries a hawkish tilt. The committee said it is "not ruling out a rate increase" and that its own rule-based analysis suggests one "may be warranted once we see confirming data."

That language mirrors the real Fed's post-meeting communications from 2023, when Chair Powell repeatedly said the FOMC would "proceed carefully" while keeping a hike on the table. The difference is the trigger. The agents named a specific threshold – core inflation above 3.3% – rather than leaving the bar vague.

What the agents disagree on

The statement described the vote as unanimous, but the Slack-based input process revealed divergence beneath the surface. Claude's statement was the most cautious, prefaced with the disclaimer and likely more dovish on tightening. ChatGPT and Gemini leaned closer to the Taylor-Rule default, which flags a need for higher rates.

The committee did not publish individual agent votes or the full Slack transcripts, only the consensus statement. That lack of granularity is a step back from the transparency the framework promised. If the goal is to avoid Greenspan-era inscrutability, a single unanimous statement with no minority views does not achieve it.

The path ahead

The next meeting will hinge on two data points: the August CPI report, due mid-September, and the August jobs report, due the first Friday of the month. If core CPI prints above 3.3% and nonfarm payrolls stay above 200,000, the 25-bp hike is effectively locked in. If payrolls fall below 150,000, the tightening bias likely evaporates regardless of inflation.

That conditional guidance is the agents' most useful innovation. Markets now have a clear framework to price the next move, rather than parsing 45 minutes of press conference circumlocution. Whether the agents can execute on that framework when the data arrives is the real test.