Euangelion — original explanatory figure
Created 9 October 2026, America/Los_Angeles.
Original paper SVG composition and responsive data.
Original Euangelion composition from cited facts; no third-party artwork reproduced.
Source basis: https://deploymentsafety.openai.com/gpt-6-october
Original factual source/check timestamps are unchanged by this stylistic revision.
No stated finding is not proven safety. Vendor tests exclude system-level safeguards and are not an independent replication or the separate teen-product assessment.
OpenAI reports statistically significant regression against GPT-5.6 in standard self-harm tests for October GPT-6 Sol and Luna, and gore and sexual-content tests for Luna. Filled gold markers denote stated regressions; hollow markers mean no such significant regression was stated, not proven safety.