Preliminary research
Reverse CAPTCHA: Evaluating LLM Susceptibility to Invisible Unicode Instruction Injection
AI-collected research leads through 2 October 2026, including bounded month-by-month reviews of selected social and community sources from January through September. Unranked, incomplete, not community-vetted, and subject to change.
Evaluates invisible Unicode encodings as prompt-injection carriers across several language models and tool-use configurations. The experiment finds that tool access sharply increases compliance and records model-specific differences across zero-width and variation-selector payloads.
Record
- Researcher
- Marcus Graves
- Published by
- Moltwire
- Date
In the archive
Related sources
Tags
This page is the archive's own catalogue record. The research is the work of Marcus Graves, first published at the original source. Preserved copies are kept so the citation survives its host.