On Tuesday, OpenAI announced that the company would pay $20,000 to people who find bugs in its AI systems.
The company has introduced a “Bug Bounty Program”. In which people will get $200 to $20,000 depending on the size of the bug they discover
The program is working in collaboration with Bugcrowd and a crowd-secured security platform.
OpenAI, the creator of chatGPT, has announced that the company will pay up to $20,000 to users reporting vulnerabilities in the system. The company has named it a “Bug Bounty Program”, which will offer users a reward depending on the bug’s severity.
According to the company, the reward starts from $200 and will go up to $20,000 per vulnerability.
OpenAI invited researchers to assess specific features of chatGPT, including the architecture for OpenAI systems to exchange information with other external applications.
However, it’s essential to mention that the assessment program does not include any inaccurate or damaging content produced by OpenAI systems.
Moreover, the company invites people from different fields to discover “bugs” in its AI system for sizeable cash rewards.
According to the company, “Security is essential to OpenAI’s mission. We appreciate the contributions of ethical hackers who help us uphold high privacy and security standards for our users and technology”.
In addition, the complaint also stated that “the policy outlines our definition of good faith regarding the discovery and reporting of vulnerabilities, and clarified what you can expect from us in return”.
Moreover, the company also said that “the initial priority rating for most findings will use the Bugcrowd Vulnerability Rating Taxonomy. However, Vulnerability priority and reward may be modified based on the likelihood of impact at OpenAI’s sole direction”.
OpenAI realized there is a need for such a program because it believes that “transparency and collaboration” are the critical factors to success.
OpenAI is collaborating with Bugcrowd, a platform that links businesses to a crowd of security researchers. To manage the submission and rewards process.
According to OpenAI, new submissions are accepted or rejected within two hours. Though, seven vulnerabilities are already rewarded.
The Bugcrowd page also outlines multiple issues which will not get any reward, including “Getting the model to say bad things to you” and “Getting the model to write malicious code for you.”
Greg Brockman, the president and co-founder of OpenAI, made a tweet that referred to the beginning of such kinds of programs.
Democratized red teaming is one reason we deploy these models. Anticipating that over time the stakes will go up a *lot* over time, and having models that are robust to great adversarial pressure will be critical. Also considering starting a bounty program/network of red-teamers! https://t.co/9QfmXQi9iM
— Greg Brockman (@gdb) March 16, 2023
Last month, the company launched a new model, chatGPT-4. The system is very creative, collaborative and accurate compared to its previous version.
ChatGPT is an intelligent tool that leverages deep learning to produce human-like results for conversational text inputs.
ChatGPT-4 is the fourth iteration of the chatbot and works explicitly on vast amounts of digital data to acquire the ability to produce accurate results like a human.
Currently, the model is accessible to chatGPT Plus Subscribers who pay $20/month. Besides, if you are using the new Bing, you can interact with chatGPT-4.
According to Microsoft, Bing is now using OpenAI’s GPT-4 model, the same technology that powers chatGPT-4.