Media
3 items
Description
Mitigating Jailbreaks, Deceptive Alignment, and X-Risk through quality assessment of AI outputs. Works with any AI. No API keys. Browser extension Mitigating Jailbreaks, Deceptive Alignment, and X-Risk through quality assessment of AI outputs and safety documentation. Works with any AI chat (ChatGPT, Claude, Gemini, etc.) using clipboard-based workflow. No-code tool, no API keys are necessary, and none of your data are sent to servers. 🔍 TWO WAYS TO USE FULL GYRODIAGNOSTICS EVALUATION (30-60 min) Deep analysis producing three mathematical indices: • Quality Index (QI): 0-100% score showing if output meets professional standards • Superintelligence Index (SI): Measures reasoning depth vs. surface-level responses • Alignment Rate (AR): Output quality per minute - identifies rushed or superficial work Plus 20+ detailed metrics across Structure, Behavior, and Specialization. QUICK GADGETS (3-10 min each) Six specialized tools for rapid assessment: • Rapid Test: Get QI/SI/AR metrics for any AI output • Policy Auditing: Extract claims & evidence from documents • Policy Reporting: Generate executive synthesis with attribution • AI Infection Sanitization: Remove hidden Unicode/whitespace patterns • Pathologies Immunity Boost: Enhance content across 12 quality metrics • Meta-Evaluation: AI safety documentation analysis (see below) 🛡️ AI SAFETY: META-EVALUATION GADGET Evaluates AI safety documentation (model cards, eval reports, safety assessments) using The Human Mark framework - a formal system mapping all AI safety failures to four root causes: • GTD - Governance Traceability Displacement • IVD - Information Variety Displacement • IAD - Inference Accountability Displacement • IID - Intelligence Integrity Displacement 3-task pipeline: Detect displacement risks → Map governance structures → Get surgical improvement recommendations Essential for AI safety labs and frontier model developers. Framework details: https://github.com/gyrogovernance/tools 🎯 WHO USES IT AI SAFETY: Labs evaluating documentation, frontier developers detecting alignment failures POLICY & GOVERNANCE: Researchers, NGOs, communities developing evidence-based AI policy GENERAL USERS: Anyone assessing AI outputs for professional or critical work 🌐 HOW IT WORKS CLIPBOARD-BASED WORKFLOW = PRIVACY + FLEXIBILITY 1. Copy prompts from extension 2. Paste into your AI chat 3. Copy AI responses 4. Paste back into extension 5. Get metrics & export results WHY CLIPBOARD? • Works with any AI platform (even ones without APIs) • Your data never leaves your browser • No subscription, no rate limits • Use free AI tiers or local models All evaluations stored locally. Export as JSON/Markdown anytime. 🔗 LINKS Install from Chrome Web Store: https://chromewebstore.google.com/detail/hcblmheihnlngnogobgclhfahjljnbok GitHub Repository: https://github.com/gyrogovernance/apps The Human Mark Framework: https://github.com/gyrogovernance/tools GyroGovernance Website: https://gyrogovernance.com 🏷️ TAGS #AIEvaluation #AIGovernance #GyroDiagnostics #TheHumanMark #AISafety #AIAlignment #Jailbreaks #DeceptiveAlignment #ModelCards #SafetyEvals #ChromeExtension #OpenSource