Sunday, 19 July 2026 · World
USD/EUR 0.8745 USD/GBP 0.7438 USD/JPY 162.4 USD/CNY 6.789 All rates →
RSS
EUROS The World Financial Report
Nº 8 Sunday, 19 July 2026 · World Edition
LATEST
Front Page

AI agents gain payment powers at Visa, Robinhood

EUROS Newsroom · 6h ago · 1 min read
AI agents gain payment powers at Visa, Robinhood

Visa and Robinhood are deploying artificial intelligence agents to execute trades and process payments, a structural shift that introduces new vulnerabilities for financial fraud.

Robinhood now utilizes AI agents capable of executing credit card purchases and investment trades on behalf of its users. Separately, Visa has collaborated with OpenAI to launch secure payments via agentic commerce, a move the payments giant said enables "seamless and trusted payments across OpenAI."

These developments represent a tangible shift in digital finance, moving AI from an analytical or advisory role into direct transaction execution. For payment processors and brokerage platforms, integrating autonomous agents promises to reduce friction and drive higher transaction volumes. However, handing financial credentials and execution authority to AI models introduces distinct systemic risks to the consumer payments ecosystem.

The security implications are a growing concern for market professionals. Fraudsters are adapting to this new infrastructure, using AI to target victims more efficiently than traditional methods. Ian Bednowitz, general manager at LifeLock, noted that the threat landscape has evolved far past the era of isolated hackers sending mass phishing emails. By leveraging AI, malicious actors can now orchestrate identity theft and fraud at a faster pace and lower cost.

The vulnerability largely stems from user interaction rather than flaws in the underlying technology. Jennifer White, managing director of banking and payments intelligence at JD Power, emphasized that the primary risk is how easily consumers can be manipulated into sharing sensitive data or taking actions that bypass established security safeguards.

One specific mechanism alarming cybersecurity experts is indirect prompt injection. In this scenario, scammers conceal malicious instructions within websites or other data sources that an AI agent routinely accesses. If the agent processes these hidden commands, it could inadvertently authorize fraudulent transactions or leak financial credentials.

"Consumers don't need to fear AI tools," White said. "But they do need to be more intentional about what information they share with them." As Visa and Robinhood scale these autonomous capabilities, the financial sector will be forced to balance the operational efficiencies of agentic commerce with the need for entirely new defensive frameworks against automated manipulation.