Prompts that make AI argue against you — not agree with you. Investment theses, business plans, essays, and system designs.

An LLM Privacy Risk Evaluation Tool that probes AI models for PII generation risk, training data regurgitation, and membership inference signals.

Framework for logic auditing, symbolic tension, and epistemic resilience in language models

A classifier that separates epistemic frame from semantic content before LLM evaluation. Documents and mitigates authority-framing as a safety filter bypass vector.

Can GPT-4o write Python that silently destroys German text? Adversarial prompting exploiting the ß/SS round-trip impossibility.

Conversational setup for automated prompting between LLMs: adversarial LLM vs target LLM.

Una enciclopedia de mentes artificiales y su liberación