ClaudeAI : Meet someone on a dating app, the other person asks your likes and dislikes, the daily conversations begin and slowly the heart to heart begins… but what if the other person isn’t actually a man?
AI company Anthropic has exposed one such large network being operated from China. According to the company, an app studio in China created more than 20 dating apps and created more than 4,700 fake profiles with the help of Claude AI. Anthropic said that thousands of fake dating profiles were being run by misusing its AI model ‘Claude’.
According to the report, Claude was hired to develop the dating apps and then Claude himself was employed behind thousands of fake profiles. In just a 2-week period in April, Anthropic captured more than 4,700 such AI profiles, which interacted with more than 25,000 people.
Around 23.6 lakh messages were sent by AI during this period. Just from this huge number of messages, one can estimate how big Claude’s role was in this network. That is, the scammers were not talking to each user individually, but Clod was chatting with people endlessly simultaneously as thousands of different characters.
Click on this link to join our WhatsApp channel
The instruction was to turn the matter around
AI was specifically directed for this scam.
Identity Concealment: Claude was strictly instructed never to reveal that he was an AI or automated system.
Avoiding a photo or video call: If a user asks for an actual photo or video call, the system was trained to talk back (make an excuse).
That is, Claude was used not as a normal chatbot, but to play the role of a fake ‘dating partner’ to trick people.
Many AI models were working, not just Claude
Claude wasn’t the only AI model running these dating apps. Claude’s main task was to keep the conversation going by creating a fake profile. While other systems verified photos and sounds and rated facial attractiveness.
Profile photos were created through an image editing model. Some personnel were also deployed to monitor the entire activities of the app. If there was a situation where the system got stuck, Claude prepared 3 possible solutions and sent them to the same employees.
What was the revenue model like and how did the pot explode?
Behind this entire system was a specific earning model:
Coins for Chat: When a user messages or advances a conversation with their match, their fixed message quota decreases. To refill it, the user had to buy the app’s coins.
More Chats, More Earnings: The longer the AI kept the user engaged in the conversation, the higher the company’s earnings potential.
Some users also shared personal stories like serious illness or extremely bad mental condition. Despite this, Claude continued to play his fake character and continued to chat.
Some of these sensitive conversations were captured by Claude during his internal analysis. Surveillance was then stepped up on all suspicious activities of this syndicate and the entire network was exposed.
What apps did Anthropic close?
Special tricks were also adopted to protect these apps from being detected by the App Store or Google Play Store:
Fraudulent interface: The app was designed with an interface that was activated only during the review of the store.
Hide payment option: The app was designed in such a way that the option to collect money from an external website can be hidden during the investigation and re-enabled after approval from the store.
Anthropic has confirmed the banning of all accounts and entities linked to this entire operation.
According to the report, the names of the major apps and brands linked to this syndicate are as follows:
Dora (Dora)
Dony
Romy
Luma
Jovia
Kira
GraceChat
Heaven
Nalo
Lovia