Meta's Oversight Board has found the company's deepfake moderation too slow and reliant on self-disclosure, calling for a dedicated AI content standard.