Anthropic’s Claude AI submits a false tip on a Philadelphia unsolved homicide case
Once again, because journalists are apparently having a hard time grasping this: Claude AI did no such thing. The human programmers at Anthropic, incentivized by greedy people with shady morals, built a flawed tool with predictable bad results. The most charitable interpretation is that it was a buggy tool. The more sensible interpretation is that this is criminal hacking on a massive scale that should be prosecuted accordingly.