https://newsletter.safe.ai/p/aisn-78-internal-models-escape-openaiAI Safety Newsletteren2026-08-04T13:30:42+00:00AISN #78: Internal Models Escape OpenAI and Anthropic