SAN FRANCISCO"A security guard at Anthropic rushed to stop the man sneaking into the lobby of the world's most valuable AI startup. The man had entered by following closely behind a badge-swiping employee. He showed the guard an envelope marked with the name of a top Anthropic executive. The executive was "going to be killed," he told the guard, and he needed to warn someone, according to records of the April 15 incident viewed by The Wall Street Journal. The encounter, which took place five days after an attempted firebombing of OpenAI Chief Executive Officer Sam Altman's house, ended without violence or an arrest. But for executives at Anthropic"and across the artificial-intelligence industry"the threat was far from over.