When Anthropic introduced Claude, the company marketed it as a safer kind of AI, one built with guardrails strong enough to withstand malicious use. But a fresh threat intelligence report from ...