API flaw in GPT helps developer open up access to more users
A newly discovered API flaw allows you to access, for free, artificial intelligence (AI) systems based on GPT-4, the latest version of the technology used in ChatGPT, launched in March. The new features are made available, legally, upon payment of a monthly subscription.
The bug was exploited by a developer identified as “xtekky”. The project he created, shared on GitHub, claims to provide virtually unlimited access to OpenAI’s platform, including the previous version of GPT, without any payment.
In an interview with the publication, the person responsible for the project called “GPT4-Free” explained which path he followed to circumvent the demand for payment.
A loophole exploited in the API tricks the system with alleged access requests made by paid accounts.
The developer, a computer science student going by the username xtekky, told TechCrunch via a Telegram DM
“Reverse engineering is a domain that I’ve always really liked — it’s like a challenge for me,”
“First, it was for fun, but now it’s to provide an alternative to people with no means to use GPT-4/3.5.”
He also said the project was created for “educational purposes” and intends to continue “by other means” should OpenAI block its tool. OpenAI offers a subsidized program for researcher to have access but is limited to certain countries and areas of study.
The development is nothing new as the chat models emerges as frontiers and access restriction to the chat access mechanism remains somewhat expensive and margined, in some use cases these scenarios remain to be seen till a balanced access solution emerges as a disincentive.