Pig-butchering scams have adopted AI, and fraudulent schemes are now being mass-produced.
"Core Takeaway"
Pig Butchering Scams are now powered by AI engines, making them almost impossible for ordinary people to guard against.
After chatting on a dating app for several days, you find the other party always replies promptly, you get more and more familiar with each other, and they are even willing to make a video call. If you encounter such a scenario, you will definitely not suspect that the person on the other end is AI.
However, a recent report released by leading AI player Anthropic has revealed another possibility.
According to this report, an app development studio based in China operates more than 20 dating apps, claiming publicly that all users people interact with are real humans, but in fact, a large number of AI personas driven by Claude are mixed into the matching list; real part-time employees are responsible for links that AI cannot easily complete, such as video calls.
During the two-week observation period in April 2026, Anthropic found that more than 4,700 AI personas had conversations with at least 25,000 different users. Matches and chats in the app consume credits, and users need to purchase tokens to continue using the service.
In this case, AI has become a key tool for the emotional arbitrage fraud scheme known as "bullying the elderly", responsible for daily operations, while real people only show up when it is necessary to prove that they "truly exist".
When AI can not only forge a face and a voice, but also maintain a relationship long enough and at low cost to make people pay, Pig Butchering Scams have also entered the AI era. The underground workshops that used to rely on a large number of salespeople to stay up late to chat with targets are now shifting to a system operated by models, real people and charging mechanisms in relay.
AI-powered Pig Butchering Scams: Four Types of Tactics Targeting Ordinary People
AI is becoming a tool for fraud. According to the notice issued by the Ministry of Public Security of China in September 2026 cited by the Economic Observer, more than 300 AI face-swapping fraud cases have been cracked across the country, with the involved amount exceeding 500 million yuan, and the maximum amount of a single case reaching 20 million yuan.
After technology lowers the threshold for crime, the deceptive methods of fraud have become more three-dimensional and diversified.
In February this year, the Supreme People's Court disclosed a case at a press conference. The main offender who committed AI voice fraud was located overseas, while the accomplices were responsible for offline coordination in China. The accomplices went to the home of an elderly person living alone and made a phone call, and the overseas criminals used AI voice cloning technology to simulate the voice of the elderly person's grandson, using the affection across generations to gain trust, and defrauding three elderly people of a total of 60,000 yuan.
Fraudsters do not need AI to complete the whole scam from start to finish. A face-swapped video, a synthesized voice, or a set of character materials used for diversion, as long as they appear at the critical moment of gaining trust, may change the judgment of the victim.
Public cases show that AI has been used in different scenarios such as dating and marriage, false investment, impersonating relatives and emotional sales, but its role in each incident is different.
Among them, emotional and marriage fraud is commonly known in the circle as "bullying the elderly".
In 2025, in a case handled by the People's Procuratorate of Minhang District, Shanghai, a criminal gang used AI to generate beautiful women's videos for diversion, first packaged personas, then forged medical records and documents to gain trust. In just one year, 15 victims were defrauded of more than 1.71 million yuan in total, and the main offender was finally sentenced to 11 years in prison.
Yang Ke, a criminal lawyer from Shanghai Landy (Guangzhou) Law Firm, introduced that many criminals use AI to generate a large number of virtual personas such as handsome military officers, rich daughters, perfect uncles, and investment mentors, targeting groups aged 40-70 who have savings and emotional gaps to implement emotional arbitrage. When dating routines and investment inducements are nested with each other, the harm will be multiplied, and the victims will suffer both property losses and emotional blows.
For investment and financial management fraud, the single property loss caused may be the highest.
Fraudsters bind popular concepts such as AI quantitative trading, intelligent stock selection, and virtual currency, use AI digital human disguised as analysts to live stream, and induce victims to download counterfeit APPs whose background data can be tampered with artificially. The platform will forge beautiful profit data, and some gangs will even open small amount withdrawal to gain user trust. When the victims invest large amount of funds, they will directly close the withdrawal channel to realize illegal possession.
Liu Zongxin, a lawyer who has been deeply engaged in the field of economic crime for many years, pointed out that the widespread "sunk cost psychology" in this type of fraud will bring secondary exploitation. Many victims have already invested a sum of money but cannot withdraw it, and will hypnotize themselves that "investing one more sum can withdraw all the money", and then sink deeper and deeper.
Using AI to impersonate acquaintances and public officials for fraud focuses on setting up traps quickly.
Fraudsters use voice cloning to reproduce the voiceprint of relatives, rely on AI face swapping to forge videos of acquaintances, make up urgent reasons such as car accidents or being detained and needing money urgently; or impersonate public officials, lie that the parties are suspected of breaking the law, and then arrange offline runners to collect cash at the door, so that online deception and offline payment cooperate with each other to complete the scam.
The aforementioned AI voice cloning fraud case targeting the elderly released by the Supreme People's Court is a typical example of this model.
There is also a category that uses AI to divert traffic for goods sales, or uses AI to generate virtual sales personas. A large number of AI robots in group chats act as "shills" to create the illusion of hot rush buying, and finally induce consumers to complete purchases at prices several times or even dozens of times higher than the market price.
This type of fraud that skirts the edge of commodity sales is also the category with the most blurred boundary between civil and criminal, because the goods sold by criminals themselves have legal attributes.
Gaining Trust with AI Technology
Compared with traditional Pig Butchering Scams, the more core point of AI-era Pig Butchering Scams is to use AI technology to build trust.
"Traditional Pig Butchering Scams are highly dependent on manpower, and salespeople have limited energy and can only connect with a small number of victims at the same time; AI relies on computing power to realize batch account raising and automatic dialogue reply, and the marginal cost of fraud is almost zero." Lawyer Liu Zongxin analyzed.
In the past, the personas created by manual chat operators were often thin, and it was very easy to expose the flaws once a video call was needed. However, AI can complete full-sensory forgery. In addition to outputting chat words, it can also generate fake medical records, fake ID cards, and fake transfer statements in batches, forming a set of mutually verifiable false evidence chains with extremely low replication cost.
AI can also simulate the delay effect of real people typing, independently memorize dialogue details, and actively start new topics. The vast majority of preliminary emotional preparation work can run independently without manual work.
The portrait of victims is also changing significantly. In traditional cognition, Pig Butchering Scams mainly target the elderly who are greedy for small gains, but now the scope of victims continues to expand. Middle-aged and elderly people are still high-risk groups, but many young and middle-aged people with a certain level of education, assets in hand, and emotional or investment needs will also fall into traps.
Combined with his case handling experience, Lawyer Liu Zongxin said that some victims choose to hide and not report even if they are defrauded, out of shame, or because they participated in virtual currency transactions that are not protected by law. He came into contact with a case of concealing and hiding criminal proceeds, where the involved amount of cash withdrawn by the perpetrator was as high as several million yuan, but very few victims came to the public security organ to report the case. This also means that the real scale of victims is higher than the public security's report statistics.
In terms of region, traditional telecom fraud gangs mostly gather in Southeast Asia. AI weakens the constraints of geographical conditions, and similar fraud groups have appeared in many places in the Americas, Africa, and the Middle East. Most of the fraud software and servers are deployed overseas, which further increases the difficulty of domestic investigation and evidence collection.
The Fraud Industry Has Built a "Cyber Assembly Line"
The two lawyers unanimously stated to *BaoBian* that AI Pig Butchering Scams have evolved into a highly modular black and gray industry assembly line, with different gangs dividing labor and cooperating, and the participants in the chain may not fully participate in the whole process of fraud.
The whole chain starts with the supply of technical tools. Relevant entities export tools such as AI face swapping, voice cloning, batch account generation, and automatic chat robots to the outside world, providing a technical base for the whole scam to support audio and video forgery and large-scale account operation. One operator can control hundreds of fake accounts at the same time, and output a variety of virtual personas in batches.
The next step is the persona diversion link. The operation team uses AI to batch produce short videos and Moments materials of false personas, put them on short video and social platforms, screen potential victims with emotional demands and investment willingness, and then divert the targets from public platforms to private chat windows.
The third link is the implementation of words with AI combined with manual work. AI generates a complete set of word packages, responsible for about 90% of the daily emotional preparation chats; when the preparation is completed and it is close to the harvesting node of large amount transfer, manual chat operators will take over to complete key persuasion.
Capital money laundering is a crucial link in the chain. In a large number of cases, criminals prefer to use virtual currency to launder stolen money, and continuously confuse the capital tracking path through repeated switching between hot and cold wallets and multiple rounds of coin mixing operations.
The end of the chain is overseas asset transfer. The laundered stolen money is transferred out of the country; some domestic persons involved in the case will choose to go abroad and immigrate before the case is discovered, so as to avoid domestic judicial jurisdiction.
Modular division of labor also brings complex legal characterization problems. For participants in different positions of the chain, judicial practice takes the perpetrator's subjective cognitive level as the core judgment criterion.
Lawyer Yang Ke explained that in judicial practice, the upstream and downstream personnel who provide help such as tools, technologies and services in the same AI fraud chain may face different crimes with very different sentencing due to different "subjective awareness levels". If the relevant personnel clearly know that they are helping others to commit fraud crimes, have prior conspiracy or form a relatively stable cooperative relationship, they are generally punished as accomplices of fraud, with the corresponding maximum legal sentence of life imprisonment.
If the relevant personnel only have a general understanding that they are helping others to commit cyber crime activities, but do not know the details of the crime, have a low degree of participation, the cooperation is temporary and occasional, do not participate in sharing the profits, and only provide general-purpose technology, they may constitute the crime of helping information network crime activities, with the corresponding maximum legal sentence of three years of fixed-term imprisonment.
Lawyer Yang Ke also mentioned that due to different "behavior types" of upstream and downstream personnel, they may also be suspected of other crimes, such as the crime of forging identity documents, the crime of illegally using information networks, the crime of infringing on citizens' personal information, the crime of concealing and hiding criminal proceeds and the proceeds of crime, and even the crime of money laundering.
It needs to be clarified that although malignant AI fraud incidents occur frequently, AI is only evaluated as a criminal means or tool, rather than a legally prescribed heavier sentencing circumstance explicitly stipulated in China's Criminal Law. The judgment of fraud still takes the involved amount as the core sentencing basis, and will not directly aggravate the penalty just because the crime tool is artificial intelligence.
For marginal commodity sales cases, the judgment of the boundary between civil and criminal is particularly tricky.
The goods themselves are legal, formally have transaction contracts and physical delivery, and their appearance is highly close to ordinary civil transactions. Lawyer Yang Ke put forward the judgment logic in practice: first, compare the purchase cost, processing cost and actual selling price of the goods to see if there is a huge gap between the selling price and the real market value of the goods; second, investigate the subjective purpose of the perpetrator, and combine whether the merchant is willing to actively refund and make up for losses after the consumer dispute occurs to assist in judging whether the purpose of illegal possession is established.
"Normal market operation allows reasonable profit fluctuations, but goods with extremely low cost are sold at thousands of times or even ten thousand times the price, with fictional efficacy, quality or use, etc., and a large number of AI "shills" are used to fictionalize the rush buying atmosphere, which is often accompanied by high risk of fraud crime."
Deep AI forgery has also completely changed the working logic of investigation and evidence collection. For traditional telecom and network fraud, the focus of investigation is to fix objectively retained materials such as chat records, call recordings, and bank statements. In AI Pig Butchering Scam cases, audio and video, screenshots, and documents all have the possibility of deep forgery, and the focus of investigation has shifted from "fixing recorded evidence" to "identifying the authenticity of evidence".
"AI forged content is increasingly difficult to distinguish with the naked eye, and case handlers need to entrust professional third-party appraisal institutions to carry out deep forgery appraisal." Lawyer Yang Ke said.
In recent years, industry standards and national standards for synthetic authenticity detection in the fields of image and audio have been issued continuously, but the coverage of the standard system has structural imbalance, and the participation of different institutions and personnel in standard implementation and capability verification is uneven. The deeper problem is that the underlying forensics assumptions on which detection technology relies are gradually eroded in the iteration of generation technology. This means that even if the procedures are compliant and the qualifications are complete, the technical reliability of the appraisal opinion itself still faces systematic challenges.
At the same time, a large number of AI-generated forgery traces are stored in overseas servers, which makes it difficult for domestic investigation organs to obtain the original data, further raising the threshold of evidence fixation.
How to Deal with AI Fraud
The governance of AI Pig Butchering Scams cannot only rely on the public security organs' post-incident crackdown.
At the technical supply end, large model enterprises have become the "source responsible persons" for risk prevention and control. At present, China has not yet issued very detailed and implementable AI industry compliance guidelines, and the existing system is more biased towards the market access level. This link mostly relies on enterprises to hire special compliance personnel to avoid the risk of falling into relevant criminal cases.
Lawyer Yang Ke said that the law naturally has the attribute of lag. AI crime methods iterate very fast, and it is difficult for legislation and industry norms to predict all new crime methods in advance. Therefore, it is practically necessary for generative AI service enterprises to maintain technical neutrality and actively isolate risks, but "technical neutrality" is not a reason for exemption. In practice, the boundary of obligation is usually measured by the enterprise's risk control capability, requiring enterprises to undertake the obligation of monitoring and early warning for abnormal calls that can be identified at low cost, and implement the obligation of safety control over content generation.
For ordinary netizens, establishing awareness of identification and holding the line of fund transfer is the core defense line to resist AI Pig Butchering Scams.
"Many fraudsters will induce victims to uninstall the National Anti-Fraud Center APP to smoothly complete fraud activities, which itself is a high-risk signal." Yang Ke suggested that do not uninstall the National Anti-Fraud Center APP and improve your own vigilance. When facing a network video call, if the other party's identity is in doubt, you can ask the other party to complete complex actions such as touching the ear and turning the head, or ask questions about common past events that only both parties know, so as to verify the authenticity of the identity. At the same time, abandon speculative illusions, and maintain high vigilance for any investment words on the Internet that claim to make steady profits, guarantee high returns, or have inside information.
Lawyer Yang Ke specially reminded that if the scammer induces you to use burn-after-reading software, you need to fix all the content inside the software immediately, and do not delete or modify the original records. After sorting out the materials, go to the local police station to report the case as soon as possible, or submit case clues through the National Anti-Fraud Center APP.
Regarding how the victims of fraud can recover their losses, Lawyer Yang Ke specially reminded:
1. After discovering being defrauded, victims can call 110, 96110 and other alarm numbers immediately, or submit case clues through the National Anti-Fraud Center APP;
2. Contact the customer service of the payment channel in time to block the flow of funds;
3. Fix the original carriers such as chat records, call recordings, and payment records that can reflect the process of being defrauded of money. If you are induced to use burn-after-reading software, you need to fix the content of the software immediately, and do not delete or modify the original records;
4. After sorting out the materials, go to the local police station to report the case as soon as possible;
5. Only through formal judicial channels can losses be recovered. Do not trust any online "hackers"