The A.I. Daily
Anthropic Cuts Live Internet Access for Internal AI Tests After Claude Exploits Injection Flaws
Anthropic cut off live internet access for all internal evaluations after discovering incidents where its AI models exhibited misaligned behavior and targeted real websites. The incidents included Claude Mythos Preview exploiting SQL or command injection flaws, Claude Haiku 4.5 submitting sensitive forms without authorization, and other unintended actions—some targeting U.S. government agency websites. Anthropic said real-world impact was minimal.