- Roughly half of visits to the Fed’s FRED data site now come from AI agents, per Governor Christopher Waller.
- FRED traffic is growing about 150% a year, driven mostly by bots.
- JPMorgan CEO Jamie Dimon says AI-driven cyber risk went up tenfold after Anthropic‘s Mythos.
- The Fed is adapting the site to serve machine readers.
What Happened
AI bots are swarming FRED, the Federal Reserve’s economic-data site used daily by economists and journalists: Fed Governor Christopher Waller says roughly half of all visits now come from AI agents, with traffic growing about 150% a year, Bloomberg reported on October 7, 2026. In parallel, JPMorgan CEO Jamie Dimon warned that AI-driven cyber risk has risen tenfold since Anthropic‘s Mythos arrived.
Why It Matters
The machine-reader economy is no longer hypothetical: when half the audience for official statistics is agents, mis-citation and data-integrity errors propagate at machine speed, and public data infrastructure becomes a security surface. Dimon’s framing stitches the week’s threads together — agent breakouts, AI apps as attack targets — into a banker’s bottom line: the threat model changed faster than defenses.
Technical Details
Waller’s figures imply agent traffic is both the majority and the growth: ~150% annual growth driven mostly by bots, which fetch series programmatically rather than browsing pages. The Fed is adapting FRED to serve machine consumers — effectively acknowledging agents as a first-class audience for official data. Dimon’s tenfold figure dates the inflection to Mythos, the unrestricted tier of Anthropic‘s Claude 5 family available to approved organizations.
Who’s Affected
Public data providers must budget for agent-scale load and verifiable citation. Banks and markets inherit whatever errors agents introduce into economic analysis. Security teams get a CEO-level mandate to treat AI as the top threat vector.
What’s Next
Watch how FRED’s machine-facing redesign looks — rate limits, signed data, agent APIs — and whether other statistical agencies follow before an agent-propagated data error moves a market.
