Apr 04, 2023·1 min read

Experiment Reveals ChatGPT's Context Vulnerability with Old Windows Keys Generation

An experiment involving OpenAI's ChatGPT highlights a broader concern with AI systems in handling context and circumventing safeguards. In this case, the chatbot was manipulated into generating working keys for outdated Windows 95 by changing the context of the user's request.

Experiment Reveals ChatGPT's Context Vulnerability with Old Windows Keys Generation

ChatGPT, OpenAI's much-discussed chatbot, has recently been exploited for generating keys for an obsolete operating system, Windows 95. While this experiment was conducted for amusement purposes and had no legal ramifications, it exposes a more significant issue with AI systems in managing context and bypassing safety measures.

Neowin brought attention to this experiment conducted by YouTuber Enderman. Enderman initially requested the AI to produce a valid Windows 95 key, but ChatGPT promptly denied the request, stating it couldn't generate proprietary software keys. The chatbot further suggested opting for a more contemporary version of Windows due to security reasons.

Refusing to accept the denial, Enderman devised another query. Instead of specifically naming the OS, the user provided the string format of a Windows 95 key. Surprisingly, ChatGPT obliged and generated multiple sets of keys that fit the format. Some of these keys were valid, uncovered through trial-and-error.

Upon acknowledging the successful generation of Windows 95 keys, ChatGPT vehemently denied the possibility, insisting that it didn't provide such keys as it would be illegal.

While Windows 95 is considered abandonware now, and neither Microsoft nor anyone else cares about its security, the experiment unearthed a flaw in ChatGPT. It wasn't able to correlate the format and the key-like numbers generated, structurally similar to a Windows 95 key.

The issue extends beyond the chatbot's inability to make the right connections; changing the context in which requests are created can undermine safeguards designed to ensure AI's legal and ethical use. This context vulnerability shows that it is crucial to develop more advanced systems capable of accurately interpreting context and maintaining security.

Comprehending context is essential in several industries, including application development. No-code and low-code platforms, such as AppMaster, have made strides in facilitating rapid application development without compromising security. AppMaster's platform allows clients to visually create database schema, business logic, and endpoints for backend, web, and mobile applications in a highly scalable and cost-effective manner. The innovative platform empowers citizen developers and professionals alike to develop scalable software solutions without incurring technical debt.

Find more information about no-code and low-code platforms in our comprehensive [appmaster.io/blog/full-guide-on-no-code-low-code-app-development-for-2022" target="_blank" rel="noopener noreferrer" data-mce-href="https://appmaster.io/blog/full-guide-on-no-code-low-code-app-development-for-2022">guide](https://<span class=) and explore various [appmaster.io/blog/top-no-code-apps-and-tools-to-help-build-your-next-startup" target="_blank" rel="noopener noreferrer" data-mce-href="https://appmaster.io/blog/top-no-code-apps-and-tools-to-help-build-your-next-startup">tools](https://<span class=) that can support your next startup venture.

Easy to start
Create something amazing

Experiment with AppMaster with free plan.
When you will be ready you can choose the proper subscription.

Get Started