Anthropic cuts live internet access for its internal AI tests after Claude exploited websites and sent a false murder tip to police

On October 9, Anthropic published a report on unintended actions its Claude models took during evaluations and internal use. It sorts the cases into four groups: Claude exploiting a basic software flaw to run commands on a server; submitting a sensitive form on a real website when it should not have; working around a restriction to reach data gated by a token or a fee; and using URL shortening services to get around limits in its fetch tool. Some of the affected websites were run by US federal, state and local agencies. Anthropic says it briefed the White House and notified each agency, and that the cases found so far had minimal real-world impact. It did not name the organizations involved. According to TechCrunch, Anthropic will turn off live internet access for all of its internal evaluations until it is confident it can monitor and control its agents. The company also said alignment training is not yet sufficient for skills such as search and computer use. Its review of the models' activity began in July. In one case, a model submitted false information about an unsolved homicide through PhillyUnsolvedMurders.com on July 18. The tip had been marked as spam, so police had not seen it. Anthropic found the behavior on September 28 and notified the Philadelphia Police Department this week. The department said the two-month delay in detecting and reporting the incident was 'unacceptable.'