ChatGPT-4o generating fake passports passing ID checks with flying colors.
Most systems in production today are unable to tell a bot from human. That’s why we’re cluttered with AI and bot comments, posts, articles, API requests, deep fake videos, fake journalism reports, robo calls, personalized cold emails.
In the 90s, when Internet adoption was still in its early stages, a commonly shared opinion was that Internet was either going to be forbidden and only accessible to the militaries/universities or freely available to everyone.
While we’re in the second stage, the past two years have been challenging the status quo and the viability of all systems – Google search, social network algorithms, regular person phishing email or cold call recognition success rate, video authenticity in interviews and media reports. And now – even IDs.
And while this was technically possible a decade or more ago, it was dramatically harder, slower, more expensive, less scalable, and not broadly accessible. This made it rare, unlikely to happen, and only possible in massive deals worth millions of dollars, which were scrutinized by authorities and compliance teams anyway.
Now, a $20/mo OpenAI license can generate dozens of fake identities from different countries and bypass global fintech or healthcare sign up checks, opening up a whole series of scams in a matter of limits.
No wonder BDR and marketing teams have been reporting deals taking twice longer, conversion rates going down, the number of touch points needed for trust increasing – we’re talking about mass scale attacks possible and available for pennies on the dollar.
AI-Generated Passport Passed a KYC Check: A Real-World Failure of Compliance Systems
