Anthropic might have fear-marketed itself into the AI export ban. 5 in every 1000 words used by Anthropic in 2026 are related to risk, regulation, or restrictions. this is 8x more than Sam Altman. Anthropic vs OpenAI use of each word in their statements: > “risk”: 336 vs 30 > “safeguard”: 121 vs 33 > “vulnerability”: 128 vs 10