"'AI scams' are spreading"…Humanity Protocol warns of risks on 'dating apps'

The decentralized identity-verification project Humanity Protocol has issued a warning about the risks of AI-driven scams on dating apps such as Tinder.

Humanity Protocol said on the 3rd that it ran an AI-based social experiment on Tinder from October through December last year. The core of the experiment was to demonstrate how AI can generate fake profiles and manipulate real users by bypassing know-your-customer (KYC) checks on major dating platforms. "The results clearly showed how vulnerabilities in existing identity-verification systems can be exploited in the era of generative AI," the project said.

Six members of the Humanity Protocol team participated in the experiment. First, the team used publicly available AI tools—such as ChatGPT, Nano Banana and Reve AI—to create a total of four fake Tinder profiles. The profiles included both photos and written bios.

The team also used AI for chats. Specifically, it used TinderGPT, which is publicly available on the open-source platform GitHub, to automatically carry on more than 100 conversations simultaneously with real users within the app. The test accounts were created in several countries, including Thailand, Spain and Portugal, but their locations were all switched to Portugal via Tinder Gold.

During the experiment period, the fake accounts interacted with a total of 296 real Tinder users. Of those, 40 agreed to go on an in-person date with the fake accounts.

Users who communicated with the fake accounts met the experiment team at a restaurant in Lisbon, Portugal. They were briefed on the experiment at the restaurant and were treated to dinner by the team. A Humanity Protocol representative said the team explained on site that it had no intention of causing financial or emotional harm through the experiment, and that its purpose was to highlight structural vulnerabilities across the system.

Terence Kwok, founder of Humanity Protocol, stressed that "this experiment was not a prank to deceive people." He said it was intended to "stress-test the systems we rely on every day and show how easily trust can be engineered when AI is involved." He added, "If a small team can do this, imagine what an organized group with malicious intent could do."

Humanity Protocol said the experiment underscores the need to improve existing identity-verification systems beyond dating apps. In particular, it argued that current identity checks are showing clear limitations because they were designed for a world without generative AI. "With AI capable of producing hyperrealistic images, voices and behaviors, there are limits to verifying identity through photo verification or ID uploads alone," it said.

Losses from romance scams are continuing to rise. According to the U.S. Federal Trade Commission (FTC), as far back as 2022—four years ago—romance-scam losses in the U.S. had already exceeded $1.3 billion (about 1.9 trillion won), the largest among all categories of consumer fraud.

Humanity Protocol noted that as AI capabilities accelerate, the line between harmless experimentation and large-scale fraud is blurring rapidly. It added that without a stronger framework that can prove "humanity" beyond simple identity checks, digital platforms are likely to become breeding grounds for the next generation of AI-driven scams.