Skip to main content

ChatGPT creator launches bug bounty program with cash rewards

ChatGPT isn’t quite so clever yet that it can find its own flaws, so its creator is turning to humans for help.

OpenAI unveiled a bug bounty program on Tuesday, encouraging people to locate and report vulnerabilities and bugs in its artificial intelligence systems, such as ChatGPT and GPT-4.

In a post on its website outlining details of the program, OpenAI said that rewards for reports will range from $200 for low-severity findings to up to $20,000 for what it called “exceptional discoveries.”

The Microsoft-backed company said that its ambition is to create AI systems that “benefit everyone,” adding: “To that end, we invest heavily in research and engineering to ensure our AI systems are safe and secure. However, as with any complex technology, we understand that vulnerabilities and flaws can emerge.”

Addressing security researchers interested in getting involved in the program, OpenAI said it recognized “the critical importance of security and view it as a collaborative effort. By sharing your findings, you will play a crucial role in making our technology safer for everyone.”

With more and more people taking ChatGPT and other OpenAI products for a spin, the company is keen to quickly track down any potential issues to ensure the systems run smoothly and to prevent any weaknesses from being exploited for nefarious purposes. OpenAI therefore hopes that by engaging with the tech community it can resolve any issues before they become more serious problems.

The California-based company has already had one scare where a flaw exposed the titles of some users’ conversations when they should have stayed private.

Sam Altman, CEO of OpenAI, said after the incident last month that he considered the privacy mishap a “significant issue,” adding: “We feel awful about this.” It’s now been fixed.

The blunder became a bigger problem for OpenAI when Italy expressed serious concerns over the privacy breach and decided to ban ChatGPT while it carries out a thorough investigation. The Italian authorities are also demanding details of measures OpenAI intends to take to prevent it from happening again.

Editors' Recommendations

Trevor Mogg
Contributing Editor
Not so many moons ago, Trevor moved from one tea-loving island nation that drives on the left (Britain) to another (Japan)…
The best ChatGPT Chrome extensions to bring AI to your browser
GPT EZ themes.

Want ChatGPT closer to hand? One of the best ways to have quick and easy access to ChatGPT is through its range of Chrome extensions. They also give you better ways of using ChatGPT, including helping you to write better prompts to get improved responses, or granting ChatGPT the ability to search the internet, which provides access to more up-to-date information.

Here are the best ChatGPT Chrome extensions you can use right now.

Read more
OpenAI’s new Shap-E tool is Dall-E for 3D objects
Purple and pink-diamond on blue background by Rostislav Uzunov.

OpenAI's latest endeavor, Shap-E, is a model that allows you to generate 3D objects from text, not unlike how Dall-E can create 2D images.

According to OpenAI, Shap-E is "a conditional generative model for 3D assets. Unlike recent work on 3D generative models which produce a single output representation, Shap-E directly generates the parameters of implicit functions that can be rendered as both textured meshes and neural radiance fields."

Read more
ChatGPT tried to pass a college exam and it didn’t go well
A MacBook Pro on a desk with ChatGPT's website showing on its display.

Studying to pass your college exams can be a stressful experience, but what if ChatGPT could just write your papers for you? That’s the question posed by researchers at the U.K.’s University of Bath, and their findings might surprise you.

According to a report from the BBC, artificial intelligence (AI) tools like ChatGPT excel at certain tasks. James Fern, a lecturer at the University of Bath, noted that “Multiple choice questions, for example, [ChatGPT] will handle those very well. We definitely were not expecting it to do as well as it did ... it was getting close to 100% correct.”

Read more