A team of cybersecurity experts from Avast warns about LoveGPT, an AI tool that helps with dating scams.
Attractive profiles and stimulating conversations are important in online dating. An online tool called LoveGPT was designed to help people with both.
However, since integrating OpenAI's GPT language models, LoveGPT can also be used for automated catfishing in online dating, according to Avast research: Scammers use fake profiles and flirty messages to trick victims into revealing sensitive information such as credit card details. These attacks are not new, but they can be scaled up using large language models like OpenAI's GPT.
LoveGPT upgrades to OpenAI technology
According to Avast, the software behind LoveGPT has been around for more than a decade. It wasn't until March 2023 that language models from OpenAI were integrated. This, Avast says, has made LoveGPT's interactions more believable, allowing it to create more convincing fake profiles and have more authentic conversations with unsuspecting users on various dating platforms.
According to Avast, LoveGPT's capabilities are far-reaching. It can bypass captchas, hide access through proxies and browser anonymization tools, create a large number of fake accounts, send and receive emails, and interact with users.
LoveGPT uses GPT-3 and GPT-3.5 to create profile descriptions and interact with users on dating sites via messages. The tool can also adapt to the typical habits of different dating platforms with appropriate prompts, such as presenting itself as a poet on Bumble or faking enthusiasm for travel on Zoosk.
Using language models makes interactions seem more human, increasing the likelihood of fooling unsuspecting users. In particular, the wide range of obfuscation tools that make LoveGPT look like a normal dating site user makes the tool suspicious. LoveGPT can be used on the following platforms:
- Ashley Madison
- Facebook Dating
- Plenty of Fish (POF)
Avast advises caution when dating online: Bots powered by large language models could take advantage of our need to socialize. You should be cautious with new online contacts and not trust them with personal information, even if they seem authentic.
I have asked OpenAI why a service like LoveGPT is allowed to access their language models and if this does not violate OpenAI's content policy. I will update the article if I get an answer.