Brokerage circles are also stirred up by "Lobster": on one side banning private installations, on the other side researching reports to find profits.

Ask AI · How do brokerages balance the efficiency and security risks of OpenClaw?

Reporter: Wang Yandan Editor: Zhao Yun

As OpenClaw (an open-source AI agent, commonly known as “lobster”) stirs up a “shrimp farming” frenzy in social circles, the brokerage industry finds itself in a strikingly contrasting “magical realism” game.

On one side, some brokerages have issued strict internal orders prohibiting the unauthorized installation of “lobster” in company environments; on the other side, several brokerage research institutions view it as a core variable in the new ecosystem of financial technology. So far, brokerages such as Founder Securities and GF Securities have published special research reports, deeply analyzing the industrial transformation and investment opportunities behind it.

Visual China Image

This parallel of “prohibition” and “research” reflects the dual considerations of the financial industry regarding security compliance and technological innovation.

Some brokerages prohibit the installation of “lobster” internally

“Our company’s intranet issued a notice prohibiting the unauthorized installation of ‘lobster.’”

A brokerage professional working in Beijing told reporters: “The financial industry’s requirements for data security are nearly stringent. The open-source nature and high system permissions of OpenClaw conflict with the existing risk control systems of brokerages. The notice stipulates that, for risk control purposes, the installation and use of any version of the OpenClaw program is strictly prohibited on company-assigned desktop computers, laptops, production and testing servers, internal shared storage devices, company-assigned work phones, and within the company network environment.”

However, this brokerage professional also stated that if employees use their personal phones or computers outside the company network, the company does not prohibit the installation or use of “lobster.” “Because the company cannot monitor private information, but currently does not advocate for employees to use it.”

Two other brokerage professionals indicated that their companies have not yet issued formal documents restricting the installation and use of “lobster,” but they do not encourage it. “Regarding technological development, especially in artificial intelligence, the company generally adopts a closely monitored approach, but it emphasizes that employees should strictly adhere to the company’s network security management system to ensure communication environment safety.”

Previously, the Ministry of Industry and Information Technology and the National Internet Emergency Center have repeatedly warned about the security risks associated with OpenClaw applications.

Recently, the Cybersecurity Threat and Vulnerability Information Sharing Platform of the Ministry of Industry and Information Technology detected that certain instances of the OpenClaw open-source AI agent pose high security risks under default or improper configurations, making them susceptible to cyberattacks, information leaks, and other security issues.

Due to the “fuzzy trust boundaries” during OpenClaw’s deployment, along with its characteristics of continuous operation, autonomous decision-making, and invoking system and external resources, the lack of effective permission control, audit mechanisms, and security hardening could lead to unauthorized operations due to command inducement, configuration flaws, or malicious takeover, resulting in a series of security risks such as information leaks and system control.

It is recommended that relevant units and users thoroughly check the exposure of public networks, permission configurations, and credential management when deploying and applying OpenClaw, close unnecessary public access, improve security mechanisms such as identity authentication, access control, data encryption, and security audits, and continuously monitor official security announcements and hardening suggestions to prevent potential cybersecurity risks.

Some brokerage teams are conducting research

In stark contrast to the internal prohibitions, many brokerage research teams are accelerating their layout of research related to OpenClaw. As of March 11, at least eight brokerages, including Founder Securities, GF Securities, Dongwu Securities, and Guojin Securities, have launched special research reports on OpenClaw, covering key areas such as multi-platform deployment, investment research scenario implementation, and reshaping computational power needs. Interestingly, some of these brokerages have internally issued documents prohibiting the installation of “lobster.”

Founder Securities’ report published at the end of February, titled “OpenClaw Empowers Financial Investment Research: 17 Efficient Application Cases Explained,” pointed out that OpenClaw, as a pioneering open-source personal AI assistant platform, is valuable because it successfully integrates the cognitive capabilities of large language models with the execution capabilities of local systems, achieving a leap from “dialogue” to “action.” The recent high market attention on OpenClaw reflects the urgent demand for practical, controllable AI tools that can truly unleash productivity.

As a result, the research team at Founder Securities conducted detailed testing on scenarios such as daily document management, email management, automated task reminders, AI-generated storybooks, in-depth research on Agents (intelligent agents), and automated browsing for information retrieval and analysis. They stated: “From the results, the overall completion efficiency and quality are very high. For many professionals engaged in proactive investment research, OpenClaw can significantly reduce the difficulty of constructing various tools, data, and quantitative stock selection strategies, and can free investors from a large amount of repetitive, rule-based work, allowing them to focus more on complex decision-making, innovative strategy development, and research. For those working in quantitative investment research, OpenClaw can also be used to complete factor research, strategy replication, portfolio construction, and testing, thereby significantly improving work efficiency.”

Zheshang Securities, in their report titled “Next-Generation Investment Research Infrastructure: From Deployment to Application of OpenClaw,” elaborated on OpenClaw’s role for different groups: “It does not replace investment research personnel but liberates humans from the manual labor of information processing, allowing them to focus more on strategy innovation, complex judgment, and value creation. For proactive investment researchers, it serves as an efficiency lever; for quantitative researchers, it is a strategy accelerator; for individual investors, it is an equalizing tool.”

It is worth mentioning that almost all brokerage research professionals, while embracing OpenClaw, have particularly issued risk warnings.

Zheshang Securities pointed out that OpenClaw’s autonomy may introduce operational risks, requiring strict configuration of permission boundaries, such as prohibiting automatic trading execution and the confidentiality of sensitive information. The outputs of large language models are random, and generated content may produce “hallucinations,” necessitating manual verification of critical data and information. Users need to exercise caution in judgment, ensuring that the data captured with AI tools does not present compliance or legal risks, and that data is collected legally and compliantly.

Dongwu Securities also warned that due to OpenClaw’s “super permissions” over the operating system, if users lack security awareness during the configuration process or connect to unreliable third-party Skill packages, it may lead to the accidental deletion or even leakage of important local files.

Daily Economic News