Could large model crimes target blockchain? Beware of emerging criminal threats
TechFlow Selected TechFlow Selected
Could large model crimes target blockchain? Beware of emerging criminal threats
The development of AI and blockchain has enabled fraud gangs and malicious hackers to obtain greater illegal gains under more covert conditions and with lower requirements.
Navigating Web3 tides with focused insightsAuthor: Xiao Sa Team
With the rapid development of artificial intelligence-generated content (AIGC) and large language models (LLMs), technology has profoundly transformed our daily lives and work methods, significantly enhancing efficiency in programming, data analysis, and content creation. However, these technological advances have also given rise to new forms of tech-enabled crime, such as Deepfake technology, which can synthesize realistic faces and voices for facial authentication. Now, with advancing technologies, fraud groups are able to carry out increasingly sophisticated, low-cost, high-profit scams—such as malicious hacking to steal funds or cryptocurrency. Additionally, the use of blockchain technology makes it even harder to trace and identify these criminal actors.
In this context, we may come across some familiar yet unfamiliar terms like "WormGPT" and "FraudGPT." These are AIGCs developed around the same time as ChatGPT, but they represent AI systems designed for malicious purposes. For instance, "WormGPT" could refer to a type of self-replicating and spreading malware that leverages technologies similar to large language models to identify and exploit system vulnerabilities. Meanwhile, "FraudGPT" might denote an AI system specifically engineered for fraudulent activities, capable of mimicking human communication patterns to deceive victims. Therefore, as technology evolves, we must remain increasingly vigilant against potential abuses and criminal behaviors.
New Crimes Spawned by Large Model Technology
Phishing emails represent a relatively common and low-cost scam method extended through this technology. Earlier this year, a Hong Kong branch of a multinational corporation fell victim to a sophisticated wire transfer fraud scheme, where hackers used cutting-edge AI-powered deepfake techniques to defraud the company of up to HK$200 million (approximately USD 25.6 million). A finance department employee received what appeared to be a phishing email from the company’s UK-based CFO, instructing them to execute a secret transaction. Although initially skeptical, the employee became convinced during a group video conference where the "CFO" and other "colleagues" appeared—all generated using Deepfake technology. This led the employee to make 15 transfers totaling HK$200 million into five different Hong Kong bank accounts. It took about a week before the company realized it had been the target of a meticulously planned scam.
Beyond scams, hackers are also riding the wave of large model advancements. Cyber attackers now recognize that controlling AI computing power is akin to holding the key to future technological development. High-value GPU clusters have thus become highly coveted targets, as these resources are essential for executing complex machine learning and deep learning tasks. And to fight AI, one must use AI: within just a year, attackers have developed dozens of distinct attack methods leveraging large models—including but not limited to phishing, malware, DDoS attacks, and supply chain attacks—all aimed at infiltrating and taking control of AI computing infrastructure. For example, earlier this year, thousands of servers at a U.S. company were breached, with the hacker group hijacking its computing cluster to mine Bitcoin, severely degrading normal user access speeds due to massive resource consumption.
Preventing New Types of Crime
Individuals and organizations must adopt practical measures to strengthen their defenses against large-model-related crimes. For individuals, raising awareness of cybersecurity is paramount. One should remain cautious toward any request asking for personal or financial information, avoid performing sensitive operations on unsecured networks, and refrain from easily sharing biometric identifiers such as facial images, fingerprints, iris scans, or voiceprints. Last week, the cryptocurrency project Worldcoin sparked privacy concerns in Hong Kong. Co-initiated by Sam Altman, one of OpenAI's founders, the project aims to provide financial services to the unbanked via iris-scanning technology. However, Hong Kong’s Office of the Privacy Commissioner found that Worldcoin violated the Personal Data (Privacy) Ordinance by failing to adequately inform participants of the purpose and risks of data collection and by not providing Chinese-language versions of its privacy notices and biometric consent forms. As a result, Worldcoin was ordered to stop using iris-scanning devices to collect citizens’ iris and facial data in Hong Kong. Individuals should handle their sensitive information more cautiously and avoid participating in projects that may infringe on privacy without fully understanding their privacy policies. At the same time, when designing and implementing systems or services involving personal data collection, enterprises must comply with laws such as the Personal Information Protection Law, ensuring transparency and regulatory compliance.
For enterprises, establishing a comprehensive cybersecurity strategy—including data encryption, access controls, and network monitoring—is crucial. Yet, as large models evolve and scam techniques become increasingly refined, detecting fraud visually will grow more difficult. Therefore, companies can consider fighting fire with fire—leveraging AI for security monitoring. For instance, adopting anti-fraud large models similar to Shanghai Telecom’s “anti-scam large model” enables real-time analysis and early warnings, allowing quick identification and response to potential threats. Regular cybersecurity training for employees to enhance their ability to recognize and respond to online scams is also a vital component of corporate defense against large-model-driven crimes.
Final Thoughts
The advancement of AI and blockchain allows fraud groups and malicious hackers to gain greater illicit profits under lower thresholds and with increased stealth. For individuals, exercising caution in the digital world can help mitigate the risk of being scammed to some extent; for enterprises, however, defending against highly precise fraud attempts and cyber intrusions becomes increasingly challenging. In practice, both individuals and organizations need to continuously monitor cybersecurity developments, regularly update and refine their defensive strategies, and when necessary, leverage AI tools for anti-fraud operations. Through these comprehensive measures, individuals and businesses can significantly improve their resilience against large-model-related crimes, reduce potential security risks, and safeguard their information assets.
Join TechFlow official community to stay tuned
Telegram:https://t.me/TechFlowDaily
X (Twitter):https://x.com/TechFlowPost
X (Twitter) EN:https://x.com/BlockFlow_News
