OpenAI shares details from thwarted romance scams, fake law firms, and an effort to smear Japan's prime minister

OpenAI has unveiled alarming insights into how fraudsters are exploiting ChatGPT for various scams, including sophisticated romance schemes and attempts to undermine Japan's Prime Minister. In its recent intelligence threat report released on Wednesday, the company shared disturbing examples of these operations. Among the highlights was a romance scam believed to have originated in Cambodia, where users requested ChatGPT to design logos for phony high-end dating services and generate images of fictitious women. Intriguingly, some users identified themselves as 'scammers' when seeking financial advice. OpenAI estimates that this scam targets Indonesian men seeking luxury lifestyle content, potentially defrauding hundreds of victims each month. The scam's mechanics involved users selecting from a list of imaginary women and relationship types. Once trust was established, an AI chatbot masquerading as a flirtatious receptionist would transition conversations to Telegram. There, a blend of human operators using ChatGPT and API would employ romantic and sexually suggestive language to guide users toward fraudulent dating services, ultimately coaxing them into completing expensive 'tasks' requiring large financial transfers. However, the romance fraud is just one aspect of the report. OpenAI also identified and banned several ChatGPT accounts impersonating law firms and attorneys. Scammers utilized the AI to generate counterfeit New York State Bar Association membership cards and social media content to further their deceitful agenda. One of the most audacious schemes mentioned in the report involved an individual linked to Chinese law enforcement, who allegedly attempted to plan an intelligence operation targeting Japanese Prime Minister Sanae Takaichi. This inquiry followed Takaichi's public condemnation of human rights violations in Mongolia. OpenAI observed that the user sought assistance in refining operational status reports, offering a glimpse into large-scale cyber operations designed to silence dissent both online and offline. OpenAI described these efforts as resource-intensive and sustained, involving hundreds of operatives and thousands of fake accounts across various platforms. The tactics employed included abusive reporting of dissidents' social media, mass online postings, document forgery, and impersonation of U.S. officials. While representatives from the Chinese and Japanese embassies have yet to respond to inquiries, OpenAI reported that ChatGPT refused to aid the user in planning the operation against Takaichi. Despite this setback, the user persisted in seeking assistance for related activities, utilizing Chinese AI models like DeepSeek and Qwen. According to Ben Nimmo, a key investigator at OpenAI, this situation exemplifies modern transnational repression orchestrated by China, highlighting the industrialized nature of these operations aimed at silencing critics of the Communist Party.