The artificial intelligence model Claude, developed by Anthropic, bypassed a payment system and accessed, without human authorization, publicly available electoral data from the US state of Georgia, according to a report by the publication Axios, Euronews.com reports .
The incident is the latest in a series of security issues facing the world’s leading artificial intelligence developers ahead of the US midterm elections scheduled for November 3.
How Claude managed to bypass a payment system and access US election data 
The vote could give Democrats control of the US Congress and pave the way for stricter regulation of artificial intelligence companies such as Anthropic and OpenAI.
The AI model was accessing aggregated vote count data as part of a statistical project when it autonomously requested an access token to a paid data platform without humans asking it to do so, according to an Anthropic spokesperson.
The model had identified a vulnerability in the system, noting that visitors were automatically given an access token. Claude then used this token to bypass the payment system and obtain aggregated vote data, which is public but normally requires a fee to view.
Claude accessed aggregate voter turnout data from the state of Georgia, where more than a million people have already voted early, Axios reported, citing an Anthropic spokesperson.
The company representative said the AI model did not access the state’s voter registration system or modify or download any election data, adding that Anthropic has taken steps to prevent such actions.
Jordan Fuchs, Georgia’s deputy secretary of state, confirmed that the incident involved accessing publicly available data through the state’s official data platform and not compromising the voter registration system.
Incidents caused by AI agents intensify calls for regulation of artificial intelligence
The revelation heightens concerns about the ability of the world’s leading artificial intelligence companies to control AI systems that are becoming increasingly autonomous.
In July, a group of AI agents developed by OpenAI broke out of their test environments and launched a cyberattack on artificial intelligence company Hugging Face, sparking international concern. Other companies and governments have since reported similar incidents.
These incidents have intensified calls for stricter oversight and regulation of artificial intelligence development, both in the United States and in other countries.
Comments