Trump backs AI safeguards for banking system, citing risks and protections

Donald Trump said artificial intelligence could “probably undermine confidence in the banking system,” yet insisted the same tools could make finance safer if they were managed correctly. The split-screen message captures the policy bind now facing Washington: push AI adoption fast, but do not let a model, a vendor or a hacker destabilize the core of the financial system.

Trump said government safeguards should exist, including the idea of a kill switch. That warning lands in a sector where officials fear less a single flashy breach than a chain reaction. Banks still run much of their infrastructure on legacy systems, and newer AI tools can be used to scale phishing, automate fraud, probe vulnerabilities and amplify cyberattacks faster than institutions can patch their defenses.

The concern has sharpened around Anthropic’s Mythos Preview, announced on April 7 and not being made generally available. Cybersecurity and financial industry experts have warned that the model could supercharge complex attacks, and that risk is especially dangerous in a banking system built on interconnected vendors, common software and old technology that was never designed for autonomous AI agents.

The urgency has already pulled top officials into the room. Jerome Powell and Scott Bessent met with major U.S. bank chiefs this week to discuss the cyber risks raised by the model, while officials from the United States, Canada and Britain also met with banking leaders over the threat. Some Wall Street banks have begun testing Mythos internally, a sign that financial firms are trying to learn from the technology even as they fear what it could do in the wrong hands.

Trump’s comments also highlight a policy contradiction inside his own administration. In March, the White House said the federal government should preserve public trust in AI while addressing AI-enabled scams and national-security concerns. But in January 2025, Trump rescinded a 2023 Biden-era executive order that had directed agencies to review AI across the economy, including financial services, and to address financial-stability risks posed by the technology.

Treasury had already flagged the danger. In a December 2024 report, the department said AI could broaden opportunities in finance while amplifying risks tied to data privacy, bias, third-party providers and cybersecurity. Treasury said it received 103 comment letters from stakeholders, underscoring how wide the stakes run from large banks to the vendors that sell them critical software.

Whether Trump’s new language marks a real regulatory turn or only a rhetorical one will depend on what follows the warning. A kill switch is a start. Binding testing rules, vendor oversight and limits on autonomous use inside core banking systems would be the test.