[ad_1]
Software vulnerabilities and bugs are a significant challenge for companies, as hackers often use them to gain unauthorized access to sensitive data. Now, in an effort to make its systems more resilient towards these bugs and vulnerabilities, OpenAI has announced a bug bounty program that will reward users for finding security vulnerabilities. The bounty program will run on the Bugcrowd platform and will offer cash rewards of up to $20,000 for exceptional discoveries and $200 for low-severity findings.
OpenAI says that the program will focus on vulnerabilities related to privacy and cybersecurity in the company’s web domains and APIs. Additionally, the company is also interested in discovering weaknesses in how ChatGPT processes user data, which will also include third-party access via its API.
Researchers can report bugs in OpenAI’s Application Programming Interface (API) and ChatGPT directly on the Bugcrowd platform. However, OpenAI has requested researchers to report model issues via a separate form unless they have a security impact. This is because model safety issues require significant research and a broader approach, and they are not discrete bugs that OpenAI can directly fix.
Limitations of the program
The bug bounty program has some limitations, as OpenAI will not accept jailbreaks for ChatGPT or text prompts intended to trick the AI program into violating its own rules. Furthermore, the company will also not accept reports of ChatGPT misbehaving or making up facts under the program.
Although OpenAI’s reward amounts may be lower than other tech giants such as Google and Apple, which can pay up to $2 million for the most severe vulnerabilities, the bug bounty program provides users with an opportunity to contribute to the improvement of the platform and protect themselves and others from potential security risks.
“We invite you to report vulnerabilities, bugs, or security flaws you discover in our systems. By sharing your findings, you will play a crucial role in making our technology safer for everyone,” says OpenAI.
[ad_2]
Source link