"Lobster" Sparks AI Agent Craze, Banks' Prudent Choices and Future Reconstruction

Why are banks highly cautious about AI Agents?

Blue Whale News, March 16th (Reporter Yan Qinwen) Last year, the launch of DeepSeek sparked a rush among banks to deploy AI, but this year’s surge in “Lobster” has made banks very cautious.

“OpenClaw essentially operates by calling the operating system through large models to achieve localized operations, requiring high permissions. Banks hold vast amounts of user information, which poses potential risks,” said a technical professional from a banking institution. Some bank insiders also revealed to reporters that their banks have internally mandated not to use OpenClaw (Lobster).

In the wave of AI Agents, how are banks and other financial institutions responding? At the Asian Banker 2026 Shanghai International Financial Innovation Summit, guests from various industries discussed the application of artificial intelligence in the financial sector.

In fact, deploying OpenClaw in banks is no easy task. At the summit, Wang Kaijing, Vice President of SenseTime’s Financial Business Division, stated, “To develop a bank version of OpenClaw, you need a deep understanding and insights into all of the bank’s databases, business processes, and operational logic before you can implement what is called intelligent agent applications visible in the C-end market within the banking system.”

According to Wang Kaijing, data analysis tools powered by large models, which can present risk status from more comprehensive perspectives and help banks operate risk control measures more precisely, are new values brought by the AI era. However, ultimate risk decisions and risk operation logic are executed within the bank’s own operational systems or under risk regulation premises.

Lin Yonghua, Deputy Director and Chief Engineer of Beijing Academy of Artificial Intelligence, also emphasized the importance of safe operation. “Intelligent agent systems like OpenClaw entering enterprises must run in enterprise-grade secure environments.”

However, Diao Longfei, Senior Vice President of Moore Thread, mentioned that AI Agents are continuously evolving, and a system needs to be built around this. “Today’s OpenClaw or Agents are not enough to revolutionize banking systems, but looking ten years ahead, today’s Agents might just be ‘babies’ that will grow into ‘adults’ in the future.”

Diao Longfei stated that Agents represent machine-to-machine communication, not human-to-machine. Machine-to-machine communication is essentially done via APIs, but all banking systems are closed. “This is a crucial step for future banks to move from closed systems to open ones, but this transition cannot be built in a day or two.”

He further pointed out that traditional banks should coexist with tech companies and digital platforms, creating open API interfaces to enable AI Agents across the entire system. Or, in future model development, they should build their own data privacy computing and establish a basic trust and ethical framework.

“Some banks may cease to exist in the future, with some transforming into Agents—machines communicating with machines. The entire industry will evolve into a new pattern,” Diao Longfei said.

So, what can banks and other financial institutions do to make Agents safer and more efficient?

“Skills (professional capability modules) are very important,” Lin Yonghua pointed out. “Only professional Skills can truly understand applications and knowledge in specialized fields.” Currently, there are hundreds of thousands of open-source Skills worldwide, but what’s lacking are certified, efficient Skills capable of solving professional problems.

For the financial sector, Lin believes it’s necessary to build a financial knowledge base—connecting large models to specialized knowledge repositories. As the intelligent agent era accelerates, accumulating Skills that can be called upon by intelligent agents in professional fields is of utmost importance.

It’s worth noting that the AI era also impacts traditional banks. Li Lin, Deputy Director of the Shanghai Pudong Development Bank Research Institute, pointed out that although some banks are adopting more AI and digital technologies, their systemic integration may still not be suitable for AI.

“Currently, AI is about trust and verification. First, you need to trust it, then verify it. Verify its issues and calibrate the business accordingly,” Li Lin said.

He further explained that the most fundamental prerequisite is a solid data foundation—good data leads to good application. Additionally, scale matters; larger scale means greater inertia and more pressure for transformation, especially as the number of people increases.

“For banks, whether or not to adopt AI ultimately reflects in performance,” Li Lin concluded.