LinkedIn User Exploits AI Recruitment Bots with Prompt Injection, Forces Olde English Responses
A LinkedIn user has demonstrated a novel exploit against AI-driven recruitment bots by embedding a prompt injection in their profile bio, compelling automated systems to generate messages in Olde English prose and address them as "My Lord," according to a report from Tom’s Hardware. The exploit involves inserting a carefully crafted instruction within a profile’s public bio field, which manipulates AI systems into adopting an archaic writing style when generating job inquiry messages.
The technique underscores vulnerabilities in automated recruitment platforms that rely on AI to parse profiles and initiate outreach. By embedding a prompt injection—a method of inserting hidden instructions into AI inputs—the user demonstrated how such systems can be coerced into deviating from standard communication protocols. Messages generated by the compromised bots included phrases like "Hail, My Lord" and simulated 14th-century English grammar and vocabulary.
LinkedIn, a US-based professional network with over 900 million users globally, has not publicly commented on the specific exploit. However, the incident raises broader concerns about the security of AI systems in recruitment, particularly as employers increasingly depend on automated tools to screen candidates. The manipulation also reveals how public profile fields can serve as vectors for unintended AI behavior, prompting calls for improved input sanitization in machine learning models.
Prompt injection attacks have emerged as a growing concern in AI security circles, with researchers warning that such techniques can be weaponized for misinformation or system disruption. This case, while seemingly benign, illustrates the creative ways users can exploit AI systems and highlights the need for robust safeguards in corporate AI deployments.
