On May 29, 2026, Ronghui Gu, CEO of blockchain security audit firm CertiK, sounded the alarm, calling the widespread adoption of AI agents an “imminent disaster.”He shared tips on how to isolate AI agents during the testing phase to prevent them from accessing sensitive personal information and digital assets. His remarks were reported in an article by Coindesk, a U.S.-based media outlet specializing in crypto assets.AI technology is evolving at a remarkable pace, and its range of applications is expanding daily. However, at the same time, security risks are becoming increasingly apparent. In particular, the possibility that autonomously operating AI agents could directly interact with users’ digital assets and confidential information is an issue that requires careful consideration.Mr. Gu’s observations offer a critically important perspective on the safe development and operation of AI agents. Protecting digital assets is always a top priority for Japanese professionals interested in crypto assets. Before AI agents become deeply integrated into our lives and financial systems, we must understand the potential risks and take appropriate measures.In this article, we will explain in detail the characteristics of AI agents underlying Mr. Gu’s warning, as well as specific approaches to protecting digital assets. We will explore what we should do now to safely build the future of AI technology and digital assets.
What Are AI Agents?
An AI agent refers to a program equipped with artificial intelligence (AI) that autonomously performs specific tasks. For example, it can gather information based on user instructions, manage schedules, or automatically execute financial transactions.
These agents utilize technologies such as machine learning (a technology that learns patterns from data to make predictions) and natural language processing (a technology that understands and generates human language). As a result, they possess the ability to assess complex situations and respond flexibly.
However, due to their high degree of autonomy, concerns have been raised about the risks posed by unexpected behavior or malfunctions. In particular, when digital assets (such as crypto assets managed by blockchain technology) or personal information are involved, the consequences could be severe.
The Background Behind CertiK CEO’s Warning
Ronghui Gu, CEO of CertiK, has issued a strong warning that the mass deployment of AI agents is an “imminent disaster.” This warning stems from concerns about the high level of autonomy possessed by AI agents and the potential security risks that accompany it.
Mr. Gu points out that even during the testing phase, AI agents have the potential to access critical personal information and digital assets. This is because, without proper isolation measures, there is a risk of unintended data leaks and asset losses.
CertiK is a company specializing in security audits of blockchain technology. Therefore, Mr. Gu’s remarks are based on expert insight that AI agents could pose a new threat to the security of digital assets.
Potential Risks of AI Agents
The potential risks of AI agents are wide-ranging. The greatest concern is unintended access to or manipulation of information due to malfunctions or bugs (program defects).
For example, one possible scenario is an AI agent mistakenly accessing a user’s digital wallet (software or hardware used to store and manage crypto assets) and unauthorizedly transferring assets. There is also a risk of highly sensitive personal information being leaked to outside parties.
Furthermore, it cannot be ruled out that malicious attackers could exploit vulnerabilities in AI agents to hijack systems, steal users’ assets, or execute fraudulent transactions. These risks increase as AI agents become more closely integrated with financial systems and personal data.
Impact on Digital Assets
The impact of AI agents on digital assets is particularly severe. Because crypto assets are irreversible—meaning transactions cannot be easily reversed once sent—errors or unauthorized access by AI agents can lead directly to financial losses.
When AI agents automatically execute trades on behalf of users, they may be unable to respond adequately to sudden market fluctuations, potentially resulting in significant losses. Furthermore, for AI agents that interact with smart contracts (contracts automatically executed on the blockchain), there are concerns regarding misinterpretation or misuse of contract terms.
Therefore, the development and operation of AI agents handling digital assets require an extremely high level of security and reliability. Users must fully understand these risks when granting AI agents access to their digital assets.
The Importance of Isolation and Safety During Testing
CEO Ronghui Gu emphasizes the importance of isolation during AI agent testing. This is an essential process for identifying and addressing potential risks during the development phase.
Testing AI agents in an isolated environment—such as a sandbox, which is a test environment disconnected from external systems—can prevent damage to actual personal information and digital assets even in the event of a malfunction or security flaw.
This isolation means strictly limiting the data and systems to which AI agents have access. Possible measures include using virtual data for testing or operating the agents on a network that is completely separate from production systems.
Specific Recommendations from Ronghui Gu
The tips shared by Mr. Gu suggest specific approaches for building a secure testing environment for AI agents. The core principle is to minimize the access privileges granted to AI agents.
Specifically, the system must be designed so that AI agents undergoing testing cannot directly access real personal information or digital assets. This includes strictly configuring access control lists (ACLs—lists that manage access permissions to system resources) and utilizing data masking (the process of replacing sensitive information with dummy data).
Furthermore, the test environment itself requires robust security measures to prevent unauthorized external access. It is essential to conduct regular security audits (a process for inspecting system vulnerabilities) to detect and correct potential weaknesses at an early stage.
Future Directions for AI Agent Development and Regulation
AI agent technology continues to evolve, and its implementation in society is an inevitable trend. However, as the warning from CertiK’s CEO indicates, safety considerations must be the top priority.
Going forward, developers of AI agents will be subject to stricter requirements to adhere to the principle of “security by design” (an approach that embeds security from the design phase). Additionally, the industry as a whole is likely to move toward establishing safe development guidelines and introducing security certifications by third-party organizations.
Furthermore, governments and regulatory authorities around the world will consider the impact of AI agents on financial systems and personal information as they explore new regulatory frameworks. Users must closely monitor these developments and assess whether the AI agents they use are trustworthy.
[Source: Original Article]
Related Articles
* We are the official distributor of RedotPay in Japan.