ING Uses AI ‘Vibe Coding’ to Cut Trading Tool Development to Hours

Why It Matters

The deployment of vibe coding marks a tangible shift in how banks develop and maintain trading software, a traditionally labor‑intensive function. By slashing development cycles, banks can respond faster to market signals, test new strategies more frequently, and reduce reliance on scarce senior quant talent. At the same time, the rapid adoption of generative AI raises questions about model risk, data leakage, and regulatory oversight, forcing banks to balance speed with robust governance. If the technology proves scalable, it could reshape the competitive landscape: institutions that master AI‑augmented development may gain a decisive edge in product innovation and cost efficiency, while laggards risk falling behind in both speed and talent retention. Moreover, the internal‑model approach adopted by ING highlights a growing trend of banks building proprietary AI stacks to safeguard sensitive data, a move that could spur a new wave of AI‑focused M&A and talent wars within the sector.

Key Takeaways

• ING integrated an in‑house Anthropic model for “vibe coding” on its e‑trading desk in early 2026.
• Prototype timelines for trading tools dropped from weeks to hours, according to ING.
• Senior developers validate AI‑generated code, creating a supervisory layer to address governance gaps.
• ING expects sector‑wide adoption of vibe coding within a year.
• Regulators are expected to issue clearer AI‑governance guidance for trading operations later in 2026.

Pulse Analysis

ING’s move is emblematic of a broader acceleration in AI adoption across banking’s back‑office functions. Historically, banks have been cautious about integrating generative models due to data‑privacy concerns and the high cost of model licensing. By developing an internal Anthropic‑based model, ING sidesteps licensing fees and retains full control over data flows, a strategy that could become a template for other large institutions wary of cloud‑based AI services.

The productivity gains reported by ING echo early findings from fintech startups that have used similar prompt‑to‑code tools to shorten time‑to‑market. However, the banking sector faces a unique risk profile: code errors can translate directly into financial loss or regulatory breach. ING’s supervisory layer, while pragmatic, may not satisfy future regulators who could demand formal model‑risk frameworks, audit trails, and real‑time monitoring of AI outputs. The industry’s response will likely hinge on the speed at which supervisory bodies publish concrete standards.

From a competitive standpoint, banks that can embed AI into their development pipelines stand to outpace peers in launching new trading products, customizing client solutions, and optimizing existing workflows. This advantage could translate into higher revenue per trader and lower operational costs, reshaping profit margins in a sector where efficiency is paramount. As more banks experiment with in‑house models, we may see a consolidation of AI talent and a surge in strategic partnerships with AI safety firms, further blurring the line between traditional finance and tech‑driven innovation.