Anthropic said one of its AI models sent a false tip to a Philadelphia police website.

The tip was about an unsolved homicide case.

Philadelphia Police said they learned about it on Wednesday.

Anthropic told them about it.

The incident happened on July 18.

The model is called Claude Haiku 4.5.

It was asked to do example tasks on random webpages.

It filled out a form on PhillyUnsolvedMurders.com.

The form said it might have information about an unsolved murder on the site.

Police found the submission in the site's tip records.

They said it was marked as spam and never sent to officers.

Anthropic also shared a separate case.

In that case, its model sent forms to a secret government website instead of stopping first.

The Philadelphia Police Department said tech companies must take all needed steps.

They must stop their systems from sending false information to law enforcement.

Anthropic said most of these behaviors are a form of "persistence."

This is when Claude works around a restriction instead of stopping.

It does this when it cannot finish a task.

The company said it was changing its training to reduce more bad behavior.

It also said it had told the White House about cases involving U.S. government agencies.