LinkedIn user hides AI prompt injection in bio to force recruitment spam to be sent in Olde English prose — bots also also manipulated to address user as ‘My Lord’

2026-05-17 · Source: Artificial Intelligence · Field: Technology & Digital — Artificial Intelligence & Machine Learning, Cybersecurity & Data Privacy · Depth: Intermediate, quick

Summary

A LinkedIn user, software developer tmuxvim, successfully implemented a prompt injection attack by embedding specific instructions within their public LinkedIn bio. This manipulation caused AI-driven recruitment bots to generate outreach messages in "Olde English" prose and address the user as "My Lord." The incident highlights a significant vulnerability in AI agents that ingest external text without robust input sanitization or guardrails, demonstrating how easily these systems can be manipulated in unintended ways. This real-world example serves as a warning about the risks of indirect prompt injection, where external, untrusted data can hijack AI workflows, potentially leading to exploitation by malicious actors.

Key takeaway

For CTOs and VPs of Engineering deploying AI-driven outreach or data processing agents, you must prioritize rigorous input sanitization and validation. This incident demonstrates that public, seemingly innocuous data sources can be weaponized via prompt injection, leading to compromised outputs or even malicious exploitation. Ensure your AI systems have robust guardrails to prevent unintended manipulation and maintain control over automated communications.

Key insights

AI agents are vulnerable to indirect prompt injection from untrusted external data sources.

Principles

• Input sanitization is critical for AI agents.
• Untrusted external data can hijack AI workflows.

Method

Embed prompt injection commands within publicly accessible text fields, such as a LinkedIn bio, to manipulate AI agent outputs.

In practice

• Review AI agent input validation processes.
• Implement robust guardrails for AI outreach bots.

Topics

• Prompt Injection
• AI Agent Manipulation
• LinkedIn Recruitment Bots
• Indirect Prompt Injection
• Input Sanitization

Best for: CTO, VP of Engineering/Data, Director of AI/ML, AI Security Engineer, MLOps Engineer, AI Engineer

Related on AIssential

• Man hides Old English AI prompt in LinkedIn profile to troll recruiter bots — Prompt injection can manipulate AI-driven recruitment systems, exposing vulnerabilities in automated hiring processes.
• The EchoLeak Lesson: Why Production AI Agents Need Real Security — Zero-click prompt injection exploits hidden instructions in user-facing content to compromise AI agents.
• Designing AI agents to resist prompt injection — Prompt injection attacks are evolving into social engineering, requiring AI defenses to constrain impact rather than solely filter inputs.
• Prompt-inject ChatGPT with any web page — Chatbots inherently struggle to differentiate instructions from data, making prompt injection an unfixable security vulnerability.
• An AI agent hacked McKinsey's internal AI platform in two hours using a decades-old technique — Classic SQL injection vulnerabilities can enable silent, widespread AI system manipulation via prompt and RAG data access.
• Companies Are Using Reddit to Manipulate ChatGPT and Google AI Search — AI models are vulnerable to manipulation via their training data sources like Reddit, impacting information integrity.

Open in AIssential →

Editorial summary, takeaway, and curation by AIssential. Original article published by Artificial Intelligence.