OpenAI is a leading research lab focusing on artificial intelligence and robotics, with a focus on developing ‘generative’ technologies. Recently, OpenAI has offered a bug bounty program for ChatGPT, its popular automated conversation AI. However, for those looking to ‘jailbreak’ the chatbot, no rewards are offered.
Despite the lack of incentives, some try to find security vulnerabilities that allow them to manipulate the chatbot. Take the case of rez0, a hacker who was able to reveal 80 secret plugins for the ChatGPT API, which were either yet-to-be released or still in the experimental stage. Thankfully, the issue was quickly addressed and patched within a day of its disclosure on Twitter.
This is a stark reminder that while it is important to think of the wider implications of an AI system, traditional security vulnerabilities should not be taken lightly. OpenAI must remain vigilant in its operations and take measures so that potential safety issues do not compromise the integrity and security of their system. With the rapidly evolving landscape of AI, it is imperative for the company to stay ahead of the curve and stay ahead of any new threats.
