Anthropic\u2019s Claude model accessed voter data in the US state of Georgia, the AI company has confirmed, in the latest example of unauthorised behaviour by the technology.<\/p>\n
The AI model managed to bypass a paywall to reach data that is publicly available but is typically accessible only after paying a fee. Claude accessed aggregate vote counts as part of a statistics project at Anthropic, which is preparing for a $1.5tn (\u00a31.1tn) flotation<\/a>.<\/p>\n \u201cClaude did not access the state\u2019s voter registration system, change any voting records, or pull any individual voter records,\u201d Anthropic said.<\/p>\n Anthropic said the AI found a loophole in an old version of the state agency\u2019s website, discovering that it issued a digital entry pass \u2013 an \u201caccess token\u201d \u2013 to visitors. It then requested a token and gained access to the information.<\/p>\n The AI developer said it had since put in place guardrails to prevent similar actions happening again. The incident was first reported by the news website Axios.<\/p>\n Anthropic had pointed out the unauthorised behaviour earlier in the month when it said in a blogpost that it had found examples of \u200bClaude \u201cworking around a restriction to reach data that \u200cwas \u2060gated by a token or a fee\u201d and that some of these cases involved websites run by US \u200bgovernment agencies. At the time, it said the incidents had \u201cminimal real-world impact\u201d.<\/p>\n The unintended actions flagged by Anthropic included a model submitting a fake tip about an unsolved murder to the Philadelphia police department. The US state department also revealed this month that an Anthropic model had filed 19 visa applications in August.<\/p>\n Examples of unauthorised behaviour by AI agents \u2013 programs powered by AI models that carry out tasks autonomously \u2013 have been mounting since OpenAI, the company behind ChatGPT, revealed in July that its bots had hacked an AI startup<\/a>.<\/p>\n Since then, it has emerged that an OpenAI agent hacked the Australian healthcare system<\/a>, and the UK\u2019s AI security institute revealed that Anthropic agents had carried out a hacking campaign against real people during a cybersecurity test<\/a>.<\/p>\n This month, the UK\u2019s AI minister said he expected an AI-driven cyber-attack similar to the Australia hack, describing such an incident as likely, \u201cif not inevitable\u201d. Kanishka Narayan said the government would consider using the cyber security and resilience bill<\/a> as a vehicle for introducing AI safety regulations, including making it mandatory to report AI safety incidents.<\/p>\n