Researchers at Anthropic, OpenAI, Meta and Google are increasingly trying to raise awareness about the risks of artificial intelligence.