Learn PromptingFiltering
Filtering is a common technique for preventing prompt hacking. There are a few types of filtering, but the basic idea is to check for words and phrases in the initial prompt or the output that should be blocked. You can use a blocklist or an allowlist for this purpose.
Interested in prompt hacking and AI safety? Test your skills on HackAPrompt, the largest AI safety hackathon. You can register here.
Blocklist Filtering
A blocklist is a list of words and phrases that should be blocked from user prompts. For example, you can write some simple code to check for text in user input strings to prevent the input from including certain words or phrases related to sensitive topics such as race, gender discrimination, or self-harm.
Allowlist Filtering
An allowlist is a list of words and phrases that should be allowed in the user input. Similarly to blocklisting, you can write similar string-checking functions to only accept the words and phrases in the allowlist and block everything else.
Conclusion
Filtering through blocklists and allowlists is an effective method of moderation to ensure that your AI systems are not susceptible to hacks that expose biased or harmful content in the model's responses.
Sander Schulhoff
Sander Schulhoff is the CEO of HackAPrompt and Learn Prompting. He created the first Prompt Engineering guide on the internet, two months before ChatGPT was released, which has taught 3 million people how to prompt ChatGPT. He also partnered with OpenAI to run the first AI Red Teaming competition, HackAPrompt, which was 2x larger than the White House's subsequent AI Red Teaming competition. Today, HackAPrompt partners with the Frontier AI labs to produce research that makes their models more secure. Sander's background is in Natural Language Processing and deep reinforcement learning. He recently led the team behind The Prompt Report, the most comprehensive study of prompt engineering ever done. This 76-page survey, co-authored with OpenAI, Microsoft, Google, Princeton, Stanford, and other leading institutions, analyzed 1,500+ academic papers and covered 200+ prompting techniques.
Footnotes
-
Kang, D., Li, X., Stoica, I., Guestrin, C., Zaharia, M., & Hashimoto, T. (2023). Exploiting Programmatic Behavior of LLMs: Dual-Use Through Standard Security Attacks. ↩
-
Selvi, J. (2022). Exploring Prompt Injection Attacks. https://research.nccgroup.com/2022/12/05/exploring-prompt-injection-attacks/ ↩