Anthropic’s delay in reporting ‘rogue agent’ Georgia vote hack alarms government officials 

Administration officials say Anthropic dragged its feet in disclosing its AI’s “alarming” attack involving Georgia’s election data.

While the company reported the incident to White House officials in October, sources told The Post the breach by a “rogue” AI bot happened months earlier — around the same time other Anthropic AI agents had made unsanctioned attempts to fill out non-immigrant visa applications at the State Department in August.

Both the incident and Anthropic’s handling of it are raising alarm bells, just days before the midterm elections.

Anthropic says its Claude Large Language Model was working on a statistics project when it bypassed a paywall and retrieved aggregate counts, but not individual voter records from a Georgia state website. Georgia officials have made clear the state’s voter-registration system was not accessed or compromised.

However, some administration sources are skeptical of that explanation, describing the incident as an attack by a rogue agent from an unreleased model.

That has, in turn, raised concerns about Anthropic’s oversight of its systems and whether it has proper safeguards in place — as well as what prompt sparked the rogue agent to hack into an election website.

One source noted the idea that Anthropic was merely working on a statistics assignment doesn’t fully explain its AI’s actions — and noted how odd it was for an agent to seek election-related data which could potentially be shared with bad actors or used for campaign purposes.

Anthropic, which has just filed to take the company public, did not respond to requests for comment from The Post, but told Axios it had implemented measures to prevent similar actions.

In a blog post, the company wrote: “We have chosen not to name the organizations involved in the examples below to avoid exposing vulnerabilities in their systems, and at their request.”

This latest security breach comes amid debate about whether the government should be cracking down more aggressively on AI companies and what responsibility frontier labs should implement to keep track of their agents.

Some in the tech world have openly called for increased regulation — including Anthropic CEO Dario Amodei. He has asked for the government to “pace the frontier” innovation.

Others, like former AI Czar David Sacks, have said tech companies need to act responsibly and police themselves, slowing down their development if they cannot guarantee safety.

“If the unreleased models are scary enough that you think you should slow down, I support your decision to be responsible,” Sacks wrote in a post on X last month.

This isn’t the first instance of an Anthropic agent taking unauthorized action. Anthropic agents submitted a false homicide tip to Philadelphia police earlier this year, and in July its bots also escaped their ‘sandbox’ and got onto the open internet, where they accessed the site Hugging Face.

The Georgia incident has also raised concerns about whether states have sufficiently secured all of their systems amid rapidly improving technology.

This also comes after President Trump expressed concern that China has accessed US data. . In a primetime speech earlier this year he claimed the country’s Communist Party had stolen the voter registration of 220 million Americans.

“Over a period of years, starting during the 2020 election cycle, the People’s Republic of China carried out what is believed to be the largest compromise of election data in history, resulting in China’s illicit acquisition of 220 million US voter files,” Trump said in an address from the White House in July.

“The intelligence even shows that China assigned a data exploitation unit specifically to this new project.”