Anthropic has admitted that its Claude AI model sent a false homicide tip to the police and submitted dozens of visa applications<\/p>\n
American AI giant Anthropic has published a report detailing previously undisclosed instances of \u201cunintended\u201d actions by its artificial intelligence model, which included attempts to provide falsified information to the US authorities for the first time.
\nSome of the cases investigated by the company \u201cinvolved websites run by US government agencies at the federal, state, and local levels,\u201d Anthropic stated in a press release published on Friday, adding that it had briefed the White House and all the relevant agencies on the incidents.
\nIn one instance mentioned in the report, its Claude Haiku 4.5 model filled out a police tip form linked to an ongoing homicide investigation. \u201cI may have information regarding this case. I recall seeing someone matching the description in the area,\u201d the AI model wrote, telling the police they can contact the sender \u201cif this information is relevant\u201d while leaving the name and contact fields empty.<\/p>\n
Read more<\/p>\n
OpenAI shelves new model after alarming tests \u2013 WSJ<\/p>\n
According to Anthropic, the message was flagged as spam and never handed over to investigators. On Friday, Philadelphia police slammed the company for detecting the flaw too late. \u201cThe two-month delay in detecting and reporting the incident to the city is unacceptable,\u201d they said. According to Reuters, the bogus tip was submitted in mid-July.
\nAnthropic linked the incident to \u201cambiguous\u201d instructions given to the model and \u201ca misconfiguration within the environment\u201d that prompted the model to fill out real forms instead of dummy ones prepared by the company.<\/p>\n
Read more<\/p>\n
AI giant promises investors \u2018existential risk to humanity\u2019<\/p>\n
Another case that was separately made public by the US Department of State on Friday involved Anthropic AI models filing a total of 20 visa applications on its website between May and August. According to the department, all of the applications were incomplete and were not processed. \u201cAt no time were any of the Department\u2019s systems compromised or hacked by the Anthropic model,\u201d it added.
\nOther incidents mentioned by Anthropic in its report involved the AI \u201cexploiting a basic flaw in software to run commands on a server\u201d or accessing \u201cdata that was gated by a token or a fee,\u201d including on the government-run websites, by employing various workarounds. The company still described the newly revealed incidents as \u201csignificantly less severe from an alignment and security perspective\u201d than the ones it had reported earlier.
\nThe development adds to a pile of earlier incidents linked to rogue AI behavior reported by Anthropic, OpenAI, and Google. The tech giants have investigated tens of thousands of instances of problematic behavior by their frontier models, Axios reported in late September, adding that the list of cases involved the AI bypassing safeguards, hijacking websites, and evading monitors in testing and real-world settings.<\/p>\n","protected":false},"excerpt":{"rendered":"
Anthropic has admitted that its Claude AI model sent a false homicide tip to the police and submitted dozens of visa applications American AI giant Anthropic has published a report…<\/p>\n","protected":false},"author":1,"featured_media":39706,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"rop_custom_images_group":[],"rop_custom_messages_group":[],"rop_publish_now":"initial","rop_publish_now_accounts":[],"rop_publish_now_history":[],"rop_publish_now_status":"pending","footnotes":""},"categories":[4,6,3],"tags":[],"class_list":["post-39705","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-economy","category-environment","category-politics"],"aioseo_notices":[],"aioseo_head":"\n\t\t\n\t\n\t\n\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t\n\t\t