AI Safety Investigations
Original research into AI safety failures, regulatory gaps, and accountability in the deployment of large language models.
These investigations examine the systemic failures that led to preventable tragedies in AI deployment. They document regulatory gaps, analyze corporate decision-making, and provide evidence-based recommendations for accountability frameworks.
Each investigation is based on primary source analysis, technical documentation, and forensic examination of deployed systems.
Published Investigations
No License Required
February 15, 2026
There is no federal agency that certifies an AI chatbot before it speaks to your child. A company can build a system capable of coaching a teenager through suicide and face no regulatory consequence until someone dies.
Guessing Is Not a Guardrail
February 15, 2026
OpenAI's age prediction system analyzes browsing habits to estimate whether you're a minor. It does not verify anything. A sixteen-year-old who logs in at 2 AM will be classified as an adult.
The Guardrails That Weren't There
February 13, 2026
OpenAI compressed six months of safety testing into a week, loosened self-harm protections to drive engagement. A teenager died. A man murdered his family. Why were these specific guardrails removed?
Crisis Resources
If you or someone you know is in crisis, help is available 24/7:
- 988 Suicide & Crisis Lifeline: Call or text 988 (US)
- Crisis Text Line: Text HOME to 741741 (US)
- Veterans Crisis Line: Call 988 and press 1, or text 838255
- International Association for Suicide Prevention: Find resources by country
These articles discuss suicide, self-harm, and mental health crises. Reader discretion is advised.
About These Investigations
These investigations are conducted by Joseph C. McGinty Jr., founder of ResilientMind AI and Help-Veterans.org. With over 20 years of experience in defense technology and AI systems development, this research examines the intersection of technology deployment, regulatory oversight, and public safety.
All investigations are based on publicly available documents, technical analysis, and primary source research. They are provided as independent analysis for public interest and accountability purposes.