Operation Date Bait: OpenAI Disrupts AI-Enabled Romance Scam Network

OpenAI has disrupted a semi-automated romance and task scam network, known as Operation ‘Date Bait’, by banning a cluster of ChatGPT accounts and one API customer. The operation likely originated in Cambodia and aligned with known Chinese-led criminal scam operations, defrauding hundreds of victims per month.

Scam Workflow and AI Integration

The network operated a structured, three-stage funnel reminiscent of cold-outreach sales programs, utilizing ChatGPT and API access to scale their deceptive activities.

1. The Ping: Lead Generation

Scammers used ChatGPT to generate promotional texts for a fake high-end dating and escort service called ‘Klub Romantis’. These advertisements targeted Indonesian men interested in luxury lifestyles (using keywords like ‘golf’, ‘yachts’, and ‘fine dining’) and were posted as paid social media ads. These ads led to an AI chatbot posing as a flirtatious receptionist who directed targets to Telegram via tracking URLs and promo codes.

2. The Zing: Trust Building and Tasking

Once on Telegram, the operation combined human operators with API-powered automation. The process involved:

  • Persona Development: Receptionist personas used increasingly romantic and sexually explicit language to lure targets into joining fake dating platforms called ‘LoveCode’ and ‘SexAction’.

  • Social Proof: The platforms displayed AI-generated announcements of fictitious clients winning prizes or completing ‘missions’ to create a false sense of legitimacy.

  • The Mentor: Conversations were handed over to a ‘mentor’ who used ChatGPT to generate and translate emotionally manipulative messages. Victims were pressured to make increasingly large payments via bank transfers or digital wallets for fictitious reasons, such as purchasing VIP cards or paying hotel deposits.

3. The Sting: Final Extraction

In the final stage, scammers attempted to collect a large final payment, referred to as the ‘kill’. This involved inventing additional fees like ‘verification deposits’ or ‘compensation settlements’. Once the maximum possible amount was extracted, the victims were blocked on Telegram.

Operational Support and Business Management

Beyond direct victim interaction, the network used OpenAI models to manage their criminal enterprise. This included:

  • Technical Assistance: Using ChatGPT for guidance on integrating the OpenAI API and updating the fake dating websites.
  • Financial Analysis: Analyzing financial accounts and generating daily reports.
  • Performance Tracking: Tracking performance across departments such as the ‘Lead Generation’, ‘Reception Team’, and ‘Supervisor Team’.
  • Internal Communication: Translating messages between Chinese-speaking supervisors and Indonesian-speaking workers.
  • Target Valuation: Generating reports that assigned each victim a ‘kill’ value, representing the estimated amount to be extracted during the final stage.

Impact and Enforcement

Based on the scammers' own inputs, the network may have been interacting with hundreds of targets simultaneously and generating thousands of dollars daily. OpenAI notes that these claims cannot be independently verified.

OpenAI's policies strictly prohibit the use of its tools for fraud or scams. The company stated its commitment to collaborating with industry peers and authorities to disrupt adversarial behaviors abusing its services.

Sources