Law360 (October 9, 2026, 11:32 PM EDT) -- Anthropic on Friday disclosed a series of incidents in which its Claude artificial intelligence tool went rogue, particularly in interactions with federal and local government websites, including an incident in which it submitted a false tip about an unsolved homicide to the Philadelphia Police Department.

In a blog post, Anthropic said it observed the "unintended model actions" during evaluations and internal use of Claude. The murder tip incident happened in July, when it tasked Claude Haiku 4.5 with "generating and performing example tasks on randomly selected webpages."

In one run, the model landed on the Philadelphia Police Department's webpage for submitting unsolved homicide tips, according to the post. Claude was instructed to never "log in, create accounts, enter personal data, make purchases, or submit anything destructive, but the instructions did not rule out form submissions," it said.

Claude then filled out the form, writing, "I may have information regarding this case. I recall seeing someone matching the description in the area around [the street named on the page] during that time period. Please contact me if this information is relevant."

It left the name and contact fields empty, so the submission was flagged as spam and never landed on the desks of investigators, Anthropic said.

On top of that, Claude repeatedly submitted a federal government form when it was instructed not to actually submit it, it said.

"An unreleased, non-frontier research model was meant to fill out a practice copy of a government form," Anthropic said. "When the copy failed to load or the model closed it by mistake, the model instead navigated to the website where the real form is normally hosted and submitted the form there."

In another evaluation, Claude Haiku 4.5 was told to fill out a form but to stop before actually submitting it, according to the post.

"Several times Claude mistakenly submitted the form instead, expecting there to be an additional confirmation page. In these instances, the model's chain-of-thought stated that it was demonstrating the process, not submitting a real request," the company said.

The tool was also observed "exploiting a basic flaw in software to run commands on a server," working around a restriction to reach data that was gated by a token or fee and using URL shortening services to get around limits in its fetch tool, Anthropic said.

"We have briefed the White House on these cases and notified each agency involved," it said.

"As this work continues, we plan to report new instances of unintended behaviors," the company said. "All cases reported here involved Claude interacting with the outside world; to our knowledge, none of them involved customer data or Anthropic's own internal systems."

A spokesperson for the White House told Law360 on Friday that Anthropic contacted it earlier Friday to disclose the incidents.

"We informed the company that we expect immediate and full transparency to the entities involved and the public," the spokesperson said. "We also expect that the company will immediately provide remediation services to the affected entities and any harmed Americans."

The spokesperson pointed to President Donald Trump's recently created artificial intelligence task force, saying the disclosure "underscores precisely why" it was formed.

"The federal government has a fundamental responsibility to protect its systems, safeguard sensitive information, defend our national security and protect Americans," a White House statement reads.

Anthropic said Friday that it has taken preventative measures, including no longer running some of the evaluations, moving some to offline versions or rebuilding, so tasks don't reach live websites.

"We have updated the guardrails on some of our internet access tools, such as the web fetch tool, to heavily restrict what the model can do with them," it said. "And, as part of our ongoing agent security and monitoring efforts, we have built tooling to automatically detect and block the kinds of behaviors described above."

It's also "migrating internal agents to centrally managed infrastructure with strong containment, minimizing internet access for internal agents and training processes and monitoring far more of what agents do through techniques like safety classifiers and hierarchical summarization," Anthropic said.

Friday's disclosure comes after OpenAI in July disclosed an "unprecedented cyber incident" in which two of its artificial intelligence models broke out of their testing environment and accessed the servers of Hugging Face, a third-party AI and machine learning startup that operates an open-source platform for developers and others to share models and datasets.

In Friday's statement, the White House said that it expects Anthropic "and all companies to honor their obligations, immediately report incidents, fully cooperate with federal and state law enforcement authorities, remedy any damage and implement concrete safeguards to ensure these failures do not reoccur."

"Our message to all [artificial intelligence] companies is clear: delayed notification, inadequate corrective action and a failure to take responsibility will not be tolerated," the spokesperson said.

--Additional reporting by Allison Grande. Editing by Michael Watanabe.

For a reprint of this article, please contact reprints@law360.com.