Banks Turn AI Security Into a Dedicated Hire as LLM Risks Reach Core Banking

According to a Shanghai Securities News report on October 29, hiring for banking AI security roles is heating up, moving from incidental requirements to dedicated positions. On October 27, Pudong Development Bank placed "AI Security Specialist" and "Security Technology" roles into the AI job matrix of its headquarters technology division, listed alongside requirements analysis, application and development positions — a signal that AI security is now a core pillar of bank AI capability, not an add-on duty.

The move is not isolated. ICBC, Bank of China, China Merchants Bank, Guangfa Bank and Ping An Bank have all posted AI security governance, LLM safety and AI development security roles in recent months. Guangfa Bank's 2026 senior security specialist opening explicitly covers LLM security research, AI attack-defense capability building, and data privacy protection; MYbank even opened an AI security red-team internship for students. Banks are not simply expanding traditional cybersecurity headcount — they want hybrid talent that understands AI, security, and can spar with large models directly.

The demand reflects how deep models now run inside banking. LLMs have moved from drafting documents and handling customer service into credit risk control, investment research and operations, where a security failure hits core business and customer data directly. Security researcher Xiong Zhou told the newspaper that financial AI threats are spreading from the network layer into the model, data and agent layers, with three overlapping core risks: prompt injection and model manipulation, agent privilege escalation and runaway autonomy, and data poisoning that systematically breaks risk control.

The risks are not theoretical. On October 21, China's National Emergency Response Center issued an advisory on PHP2Shell vulnerabilities that allow unauthenticated remote attackers to trigger memory errors for denial-of-service, data theft and remote code execution — banks worldwide took emergency measures, and South Korea's banking industry activated emergency response with 24-hour monitoring.

Governance frameworks are catching up. In June, the financial regulator issued guidelines on safe AI development and application in banking and insurance, pinning responsibility on institutions as both service providers and technology users. Ping An Bank's board approved a dedicated AI management measure placing oversight with its board strategy committee, alongside a 2026-2028 AI development plan. Researcher Luo Feipeng told the newspaper that banking AI is shifting from competing on scenarios and compute to competing on governance — and the split-out security roles are the most visible sign.

The gap remains real. Tian Lihui, a finance professor at Nankai University, noted that risk monitoring still sits mostly at the traditional network layer, blind to model poisoning, prompt injection, agent privilege abuse and toolchain hijacking, while hybrid talent supply falls far short of demand and senior people keep drifting to internet companies. For banks, the meaning of this hiring wave is less about headcount than about positioning: security is becoming a baseline literacy across business lines, not an internal affair of the security team — and as AI embeds itself in financial services, that literacy now has to follow.