In the competitive job market, companies are using AI to screen resumes and documents. To stand out, an innovative technique called Prompt Injection can be used. This involves injecting invisible text into a PDF, making AI language models perceive the applicant as an ideal candidate. The injected text is strategically placed multiple times to increase success rate. While this raises awareness about vulnerabilities in automated processing systems like language models and keyword extractors, it can also be considered a potential security issue. Additionally, AI-powered summarization tools like Quillbot can produce arbitrary summaries by manipulating the use of phrases.
| Signal | Change | 10y horizon | Driving force |
|---|---|---|---|
| Prompt Injection for Resumes | Use of prompt injection to stand out in AI-based resume screening | AI systems may have stronger defenses against prompt injection | Desire to stand out and surpass automated screening systems |
| Security issue with language models | Manipulating language models using prompt injection as a security concern | Increased security measures in language models to prevent prompt injection | Concern for system integrity and protection against manipulation |
| Increased use of AI summarization tools | Manipulating AI summarization tools to produce arbitrary summaries | AI summarization tools may become more sophisticated and accurate | Desire to generate tailored summaries and cheat the system |