Tag: ai-safety

This isn't just access control—it's OpenAI admitting they've crossed into genuinely dangerous AI territory while trying to arm defenders first.