We are immersed in an increasingly digitized world in which it is not always easy to predict which will be the next technology that will change new ways of seeing (everything seems to indicate that for these times it will be the promising one artificial intelligence). But we do have some certainties, and one of them is that there are no 100% safe systems, and this also applies to OpenAI products.
Just as old-school thieves use their skills to open locks using pliers and get their prized booty, cybercriminals do their own thing with computer systems. Entonces puede que te que te que tienen quo tienen in comón estos individuos. The answer is simple: they take advantage of physical or digital vulnerabilities that have not been addressed in time.
In response to this scenario, many companies decide to reward computer bugs to avoid them security problems with its platforms This is precisely what the company founded by Sam Altman is doing. In addition to revising their systems, there is the possibility that computer experts from all over the world can contribute.
From OpenAI’s Bug Bounty
The creators of GPT-4, among other AI models, have announced the launch of their program of rewards for errors or vulnerabilities (VRP) of ChatGPT. It is Bug Bounty has a payment scheme based on the importance of contributions. The payments range from 200 dollars for discoveries of low gravity to 20,000 dollars for “exceptional finds”.
As we say, this type of practice is increasingly common among the great technologies, and throughout history we have seen industrial giants like Apple, Google, Meta or Microsoft have offered rewards of hasta 200,000 dollars or more for errors. Eso sí, there are certain conditions that those interested in participating must fulfill at the time of putting their systems to the test.
OpenAI has published a list of problems that are outside of its reward program, among those found in the use of jailbreaks, forcing the model to return inappropriate results or getting it to write malicious code. In cualquiera de estos casos, y de otros también, no hay recompensa. If there is reward in other types of failures or vulnerabilities.
The artificial intelligence company is interested in discovering authentication problems, exposed data situationswill fail in the ChatGPT Plus payment platform, a method to circumvent Cloudflare’s protection by sending traffic to endpoints that are not protected, application blocks, problems with the API or with plugins created for OpenAI (not for third parties ).
For this Bug Bounty, those directed by Altman have associated themselves with Bugcrowd, a well-known old company from the world of rewards for failures and vulnerabilities in computer systems (although there are also others such as HackerOne and Synack). Those interested can find all the details in the OpenAI profile inside the Bugcrowd platform, which will be in charge of managing shipments and rewards.
Images: Nahel Abdul Hadi / ChatGPT screenshot
In Xataka: If ByteDance wants TikTok to continue operating in the US, it’s easy: they only have to sell it
In Xataka: I’m a computer scientist and I work reporting software failures to large technological companies