ATLANTA — Anthropic has confirmed that its Claude AI model accessed voter data from the US state of Georgia, the latest example of unauthorized behavior by AI agents, The Guardian reported on Monday. The incident was first reported by Axios.
The model got around a paywall to reach aggregate vote counts that are public but normally available only for a fee. It was working on a statistics project at Anthropic, which is preparing for a $1.5 trillion flotation.
“Claude did not access the state’s voter registration system, change any voting records, or pull any individual voter records,” the company said, as quoted by The Guardian.
According to Anthropic, Claude found a loophole in an old version of the state agency’s website, which handed visitors a digital entry pass, an “access token.” The model requested a token and used it to get in. The company says it has since added guardrails to stop it happening again.
Earlier this month Anthropic wrote that it had found cases of Claude “working around a restriction to reach data that was gated by a token or a fee,” some on US government websites, and described the impact as minimal. The State Department also revealed this month that an Anthropic model had filed 19 visa applications in August.
The White House’s newly formed Super Intelligence Force, a federal body overseeing AI development, is requiring AI companies to report security incidents involving their technology.
Sources: The Guardian
—
