TL;DR
A LinkedIn user inserted a prompt in their profile to make AI-generated recruiter messages appear in Old English. This demonstrates how AI prompts can be manipulated, raising concerns about AI security and authenticity.
A LinkedIn user known as tmuxvim inserted a prompt into their profile that caused AI-driven recruiter messages to appear in Old English, illustrating a new form of AI manipulation and raising concerns over AI security.
According to a tweet from tmuxvim, they added a prompt injection into their LinkedIn bio instructing AI scanners to address them as ‘My Lord’ and speak exclusively in Old English from the year 900 AD. As a result, recruiters messaging the user received responses in an archaic language, including a notable message beginning with ‘My Lord Arthur’ and containing text in Old English script about treasure and warriors. The user shared a screenshot of the message, which was generated by AI interpreting the prompt injection.
The incident was confirmed by tmuxvim, who explained that the prompt caused AI tools scanning their profile to produce responses in Old English, regardless of the original message content. The example message was partially transcribed using OCR, revealing a mixture of intelligible Old English and nonsensical phrases, including references to gold hoards and warriors.
Discussions online following the tweet suggest this is a demonstration of how AI prompts can be manipulated, intentionally or accidentally, to produce unexpected or humorous responses. Some community members proposed further prompt injections to generate even more absurd or controlled outputs, highlighting potential security vulnerabilities in AI systems used for recruitment and communication on platforms like LinkedIn.
Why It Matters
This incident underscores the vulnerability of AI systems to prompt injection attacks, which can manipulate responses in unpredictable ways. For platforms relying on AI to filter or generate responses, such exploits could lead to misinformation, impersonation, or security breaches. It also raises awareness about the need for better safeguards against prompt manipulation in AI-driven communication tools, especially in professional contexts like recruitment.
The Developer's Playbook for Large Language Model Security: Building Secure AI Applications
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Background
Prompt injections have been a known issue in AI safety discussions, but this incident on LinkedIn demonstrates a tangible, publicly visible example of how easily AI outputs can be altered through user-controlled prompts. The event follows growing concerns over AI security, especially as AI becomes more integrated into social media and professional networking platforms. Prior to this, similar vulnerabilities have been discussed in AI research circles, but this is among the first high-profile cases involving LinkedIn and recruitment messages.
“I put a prompt injection into my LinkedIn bio and recruiters are messaging me in Old English and calling me Lord.”
— tmuxvim
“Even though I’m old and English, it doesn’t make any sense to me. I can see there’s something about a hoard of gold in there.”
— tmuxvim
Artificial Intelligence Safety and Security (Chapman & Hall/CRC Artificial Intelligence and Robotics Series)
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What Remains Unclear
It remains unclear how widespread this type of prompt injection could become or whether platforms like LinkedIn will implement measures to prevent such manipulations. The full extent of potential security risks posed by prompt injections in professional AI applications is still being evaluated, and no official response from LinkedIn or AI providers has been announced.
AI chatbot Robot Companion and smart Interactive Toy Featuring Voice Chat, Facial Expressions, Dancing and Music, Complete with a complimentary Outfit—Making it the Ideal Gift (Black)
Companion: This desktop robot is far from an ordinary toy; it is equipped with an advanced large language…
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
What’s Next
Experts expect increased scrutiny of AI prompt security and potential development of safeguards against injection attacks. Platform providers may update their AI filtering and moderation protocols. Further demonstrations or incidents could emerge as users experiment with prompt injections in various contexts.
LinkedIn AI profile security
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Key Questions
What is a prompt injection in AI?
A prompt injection involves inserting specific instructions into an input that manipulate the AI’s output, potentially causing it to produce unintended or humorous responses.
Could this kind of manipulation affect other platforms or AI systems?
Yes, prompt injections could potentially be used to manipulate responses across various AI applications, including chatbots, virtual assistants, and social media platforms, if safeguards are not in place.
Will LinkedIn or AI providers address this vulnerability?
It is not yet clear whether LinkedIn or AI system developers will implement specific measures to prevent prompt injection exploits. Ongoing discussions about AI security are likely to influence future updates.
Is this a common problem or an isolated incident?
While prompt injections have been known in AI research, this incident is among the first high-profile cases publicly demonstrating its impact in a professional networking context.
