Visa calls for AI in cyber defence as India proposes human checks for AI payments

Visa is urging enhanced cyber defence through the integration of artificial intelligence (AI) in an "AI in the loop" model, aiming to keep pace with rapidly evolving threats. Identifying and exploiting weaknesses in payment systems now takes hours, a significant reduction from the one-and-a-half years it took in 2016, according to Subra Kumaraswamy, a Visa representative speaking at the Global Fintech Fest. This acceleration means traditional, human-led security operations can no longer effectively counter threats.

Kumaraswamy highlighted the blurring lines between cybersecurity and payment security, predicting a shift in fraud tactics from human-led activities to "agentic impersonation," where synthetic identities conduct transactions on behalf of humans. This evolution necessitates novel security solutions capable of identifying and thwarting more sophisticated threats.

Visa proposed a new metric for companies to track: "mean time to adapt" (MTTA), measuring the time from vulnerability discovery to its remediation and validation. AI can rapidly identify and chain vulnerabilities into exploitable paths, making vulnerability discovery a "yesterday's problem." Therefore, a metric like MTTA, focused on rapid response and remediation, is crucial.

Visa has open-sourced the Visa Vulnerability Agentic Harness (VVAH) tool, designed to identify security flaws and generate patches. The tool, developed in collaboration with Anthropic, utilizes advanced AI models. While VVAH can propose and even apply code changes, Visa states that findings and fixes are "LLM-generated triage candidates that require human review."

Concurrently, Indian authorities, including the Ministry of Electronics and Information Technology (MeitY), have proposed mandatory human interventions in AI-based payment systems. This approach emphasizes consumer protection, ensuring human oversight in critical payment transactions despite rapid technological advancements.