A False Tariff Post Moved $2.4 Trillion. Wall Street Wants AI to Trade Faster

U.S. stocks swung by roughly $2.4 trillion in ten minutes on April 7, 2025, after a false claim about a White House tariff pause spread from X to CNBC to Reuters, according to Dow Jones Market Data. Between 10:08 and 10:18 a.m., the S&P 500 briefly erased its losses and then reversed again — all on information that was never true.

The claim circulating on X said Kevin Hassett, director of the White House National Economic Council, had told reporters that President Donald Trump was considering a 90-day pause on tariffs for every country except China. Market-focused accounts amplified it. CNBC repeated the unconfirmed claim on air. Reuters published a report citing CNBC. Stocks took off almost immediately.

There was one problem: Hassett had not said it. After the White House called the report "fake news," CNBC corrected it and Reuters withdrew its headline, sending the market back down. The episode predates much of the current push to give AI agents more authority, but it now reads like a preview of a much larger problem.

Speed itself is old news on Wall Street. Firms have used machine-readable news for years, letting software process information and trade on it almost immediately. What is changing is how much the software does before the trade happens.

AI agents can gather information from multiple sources, interpret what they find, weigh competing signals, and — depending on the authority granted — recommend or carry out an action. That makes information quality only part of the problem. An agent must also judge whether a report is credible, what it means, whether it matters to the task, and whether the evidence justifies acting.

"Provenance isn't truth," says Nathaniel Bradley, CEO of Datavault AI.

The April episode illustrates his point. A system receiving the Reuters alert could identify Reuters as the publisher and CNBC as its cited source. Neither signal could establish whether the underlying claim about the White House was true.

What happens when accurate data misleads?

Ten months later, a very different story showed why even correct information creates its own risks. In February, Citrini Research published "The 2028 Global Intelligence Crisis," a fictional account of an economy two years out where AI had pushed unemployment to 10.2% and the S&P 500 down 38% from its October 2026 high.

Citrini told readers exactly what they were reading. "What follows is a scenario, not a prediction," the authors wrote near the top. The subtitle read: "A Thought Exercise in Financial History, from the Future."

The piece went viral as investors worried about AI's effect on software companies and white-collar employment. Reuters later reported the Citrini scenario was among several bleak AI outlooks circulating as software and financial stocks came under pressure. The problem was not provenance or labeling. It was understanding what those numbers represented before treating them as data about the real economy.

Eric Ciarla, cofounder of Firecrawl, works on the layer between a web page like Citrini's and the AI system reading it. Firecrawl converts web pages into structured information that AI applications can process, stripping navigation, ads, headers, and footers while preserving the publisher's words, the source URL, and — in the Citrini case — both the subtitle and the disclaimer.

"The reading, the weighting, and the conclusion happen in the model and the prompt," Ciarla told Fast Company. "Our part is giving them the best possible version of the page to work from."

Firecrawl has also begun going directly to information providers. Its partnership with Wikimedia Enterprise gives it access to Wikimedia data through official APIs rather than repeatedly scraping Wikipedia pages.

Even when context arrives intact, the system must still decide how much weight to give it. "Trusted sources matter, but trusting the source isn't the same as trusting the decision," says Kevin Frechette, CEO of Fairmarkit. He argues agents need company data, policies, previous decisions, approval thresholds, and clear limits on what they can do without asking a person — especially when one agent's output feeds another automated system.

How far can one mistake travel?

Jim Wetekamp, CEO of Riskonnect, calls the result a "cascade of AI decisions." One agent produces information, another interprets it, and that interpretation becomes the basis for a recommendation a third system executes. "What used to take hours or days to cascade can now happen in seconds," Wetekamp says.

He offers a concrete example: an AI misclassifies a customer account as a vendor. A second system relies on that classification to skip a required customer check while a third begins vendor onboarding. Several systems may have acted on the original error before anyone finds it.

Wetekamp says companies must decide in advance where automated processes require human approval — triggered by conflicting information, low confidence, unusual circumstances, or decisions above an agreed threshold. "Accountability can't be delegated to AI," he says.

Human review gets harder, though, when the information behind a decision has already passed through several agents.

Where did that number come from?

If one AI hands another a number, Vlad Luzin wants the receiving system to know more than the number itself. Luzin, cofounder and CTO of Band, says the message should carry who sent it, what that agent was allowed to access, and whether the figure came from a database, an original document, or another model.

"Today, in most deployments, none of that context travels with the message," he says.

Luzin is skeptical that instructing agents to verify important information will suffice. Research has found AI agents do not always follow the plans they are given, and documented incidents show agents acting against instructions they had previously agreed to follow.

"You don't make the agent smarter," Luzin says. "You make the important facts impossible to misremember and the important actions impossible to take unchecked."

Keeping such records would also make mistakes easier to investigate. If a claim has passed through five agents on different systems, finding its origin could mean piecing together separate logs — assuming they exist.

"Where that exists, tracing a claim back through five agents is a query," Luzin says. "Where it doesn't, it's forensics, and often it's simply impossible." He frames the security model in terms financial institutions already know: "Know your counterparty, keep the ledger, limit the exposure, and audit everything."

Who gets to make the call?

Judah Taub, cofounder and managing partner at Hetz Ventures, sees a larger business forming around these questions. As companies give AI systems more authority, they hand over decisions that previous generations of software left to people: which source deserves more weight, whether conflicting information requires another check, whether there is enough information to proceed. Companies are now building tools for provenance, identity, verification, permissions, monitoring, and governance.

"Capability without trust simply increases the speed at which mistakes propagate," Taub says.

That is what makes April 7 more relevant now than when it happened. The false tariff claim was corrected quickly, but markets moved before the corrections caught up. Human traders have supervisors, risk limits, compliance rules, and ultimately someone responsible for their decisions. An AI agent can be given versions of the first three; there is no obvious equivalent for the fourth. If an agent reads the information, weighs the evidence, and makes the call, the industry is still drawing the line on where human responsibility begins.