Visible prompt-like text
Instructions addressed to an AI can sit inside ordinary sections, footnotes or appended content.
Document security for recruitment
Scan CVs before they influence AI-assisted screening.
CVs, portfolios and candidate files can contain text written to influence an AI system. .mdSiren is designed to add a document review before AI-assisted summarisation or screening.
Product coming soon- OutsideCandidate file
- .mdSiren scanReview before use
- After reviewReview workflow
The recruitment problem
Candidate content can become input and instruction.
A CV is supposed to describe a candidate, but an AI tool may also respond to instruction-like text inside the file. Hidden or visible prompts can therefore distort summaries, rankings or the information presented to a reviewer.
The right boundary sits before AI-assisted processing, while the original file and the human decision remain separate.
Understand document prompt injectionCandidate-file intake
Protect the integrity of the review workflow.
Use a consistent scan across the files an AI may summarise, organise or compare.
Content a reviewer may not notice
Supported hidden or machine-readable layers may be extracted even when they are difficult to see on the page.
Downstream influence
A manipulated summary or comparison can shape what a human reviewer sees and investigates next.
A safer screening workflow
Review the file before the AI reviews the candidate.
Keep document security separate from candidate assessment and apply it consistently.
- 1Receive the CV or candidate file through the normal channel.
- 2Keep it outside AI-assisted screening while unreviewed.
- 3Run the supported .mdSiren checks.
- 4Review findings and coverage without judging the candidate.
- 5Resolve document concerns through an agreed human process.
- 6Keep screening and employment decisions human-led.
.mdSiren does not assess candidates, rank suitability or make employment decisions. It is designed to support review of the document before AI-assisted use.
Before screening
Separate document integrity from candidate merit.
The scan should answer questions about the file and its supported content—not about whether a person should progress.
- Apply the same intake process to equivalent candidate files.
- Scan the exact file that will enter the screening workflow.
- Review supported findings independently of candidate assessment.
- Check coverage and retain appropriate human oversight.
- Follow applicable recruitment, privacy and data-retention rules.
.mdSiren is coming soon. This page describes a document-security checkpoint, not a recruitment recommendation or compliance guarantee.
Related resources
Before the project
Put a document check before AI-assisted screening.
Review supported document signals first, then keep candidate assessment and employment decisions human-led.