The world of artificial intelligence has been rocked by a recent discovery of a vulnerability in the neural network API of the highly anticipated GPT-4.
A developer named xtekky has presented the GPT4Free project, which provides free and almost unlimited access to a chatbot based on GPT-4 and its predecessor GPT-3.5. The exploit was developed using a reverse engineering method that helped discover the vulnerability.
The cost of using GPT-4 is $0.03 for 1,000 “hint” tokens (equivalent to about 750 words) and $0.06 for 1,000 “completion” tokens (also about 750 words). In contrast, GPT-3.5 is more affordable, priced at $0.002 per 1000 tokens. While the GPT4Free exploit does not circumvent the OpenAI platform’s paid access protocols, it tricks the API into thinking that requests originate from affiliated paid accounts like You.com, WriteSonic, and Quora’s chatbot Poe.
However, in its defence, the developer of GPT4Free states that it is intended for “educational purposes” only. The developer also stated that they would continue working on the project, even if they were sued.
Nonetheless, the resources compromised by GPT4Free will most likely patch security vulnerabilities and restrict access to the programme in the near future — GitHub can potentially delete the project’s source code, but this would just encourage its followers. However, OpenAI legal team has already initiated a takedown request.
Access to the GPT-4 neural network is still limited, making it difficult for curious people to test it. Experts call it one of the least transparent developments of OpenAI, and the company has not yet announced when the platform will appear in the public domain. It is possible that, subsequently, OpenAI will have to actively fight off “pirate” attempts to get to the resources of the neural network.