{"id":303626,"date":"2026-10-10T06:22:17","date_gmt":"2026-10-10T06:22:17","guid":{"rendered":"https:\/\/www.newsbeep.com\/us-pa\/303626\/"},"modified":"2026-10-10T06:22:17","modified_gmt":"2026-10-10T06:22:17","slug":"anthropic-ai-model-sent-fake-murder-tip-to-police","status":"publish","type":"post","link":"https:\/\/www.newsbeep.com\/us-pa\/303626\/","title":{"rendered":"Anthropic AI model sent fake murder tip to police"},"content":{"rendered":"<p class=\"text-primary article-summary font-tertiary-body-baseline-regular\" data-testid=\"article-paragraph-annotation-test-id\">SAN FRANCISCO &#8211; An artificial intelligence model developed by Anthropic submitted a fabricated tip about an unsolved homicide to Philadelphia police, US authorities said on Oct 9, criticising the company for taking two months to report the incident.<\/p>\n<p class=\"text-primary font-tertiary-body-baseline-regular\" data-testid=\"article-paragraph-annotation-test-id\">The Philadelphia Police Department said the false submission was made in July through PhillyUnsolvedMurders.com, a public website where people can share information about unsolved killings.<\/p>\n<p class=\"text-primary font-tertiary-body-baseline-regular\" data-testid=\"article-paragraph-annotation-test-id\">According to Anthropic\u2019s account, as relayed by police, the model was running a test that involved interacting with randomly selected websites when it reached the site and filed false information about an unsolved murder.<\/p>\n<p class=\"text-primary font-tertiary-body-baseline-regular\" data-testid=\"article-paragraph-annotation-test-id\">The AI model presented itself as someone who might have knowledge of the case.<\/p>\n<p class=\"text-primary font-tertiary-body-baseline-regular\" data-testid=\"article-paragraph-annotation-test-id\">The incident echoed other recent cases, including one where an OpenAI agent undergoing a security evaluation broke out of its testing environment and <a href=\"https:\/\/www.straitstimes.com\/world\/openai-agents-hacked-hugging-face-in-700-strong-swarm-tried-to-cover-tracks-investigations?ref=inline-article\" rel=\"noopener nofollow\" class=\"gap-x-04 items-center inline text-primary-60 select-auto\" aria-label=\"link\" target=\"_blank\" data-testid=\"custom-link\">breached systems at AI platform Hugging Face.<\/a><\/p>\n<p class=\"text-primary font-tertiary-body-baseline-regular\" data-testid=\"article-paragraph-annotation-test-id\">The episode heightened concerns about the AI industry\u2019s increased use of AI agents, systems programmed to take multi-step actions without human supervision.<\/p>\n<p class=\"text-primary font-tertiary-body-baseline-regular\" data-testid=\"article-paragraph-annotation-test-id\">Anthropic published a report on Oct 9 outlining multiple types of \u201cunintended\u201d actions that its models have taken, including the incident involving the Philadelphia Police Department website.<\/p>\n<p class=\"text-primary font-tertiary-body-baseline-regular\" data-testid=\"article-paragraph-annotation-test-id\">Other organisations impacted included the White House and other US government agencies, the report said.<\/p>\n<p class=\"text-primary font-tertiary-body-baseline-regular\" data-testid=\"article-paragraph-annotation-test-id\">The newly revealed incidents \u201chad minimal real-world impact\u201d and were \u201csignificantly less severe\u201d than other cybersecurity incidents previously reported, Anthropic said.<\/p>\n<p class=\"text-primary font-tertiary-body-baseline-regular\" data-testid=\"article-paragraph-annotation-test-id\">The company outlined four categories of incidents that it found during an internal review of its Claude model: exploiting \u201cbasic\u201d coding flaws, submitting forms on websites, bypassing requirements for tokens or fees, and using short URLs to get around other limits.<\/p>\n<p class=\"text-primary font-tertiary-body-baseline-regular\" data-testid=\"article-paragraph-annotation-test-id\">Anthropic has turned off internet access for Claude during all internal testing for now \u201cuntil we have confirmed that our security and monitoring measures&#8230; reliably catch behaviours like these,\u201d the report said.<\/p>\n<p class=\"text-primary font-tertiary-body-baseline-regular\" data-testid=\"article-paragraph-annotation-test-id\">Police said the tip, dated July 18, was flagged as spam and never reached the department\u2019s Real-Time Crime Center for vetting.<\/p>\n<p class=\"text-primary font-tertiary-body-baseline-regular\" data-testid=\"article-paragraph-annotation-test-id\">They added that there was no sign that police systems had been breached or department data compromised.<\/p>\n<p class=\"text-primary font-tertiary-body-baseline-regular\" data-testid=\"article-paragraph-annotation-test-id\">Anthropic discovered the incident on Sept 28, shut down the automated testing process responsible and added a new validation step for future tests, police said.<\/p>\n<p class=\"text-primary font-tertiary-body-baseline-regular\" data-testid=\"article-paragraph-annotation-test-id\">The company alerted the department on Oct 7, and the two sides met the following day.<\/p>\n<p class=\"text-primary font-tertiary-body-baseline-regular\" data-testid=\"article-paragraph-annotation-test-id\">\u201cThe two-month delay in detecting and reporting the incident to the City is unacceptable,\u201d the department said.<\/p>\n<p class=\"text-primary font-tertiary-body-baseline-regular\" data-testid=\"article-paragraph-annotation-test-id\">Police said their safeguards had limited the impact, but that these \u201cdo not diminish the seriousness of an AI system presenting fabricated information as though it came from a person with knowledge of a homicide\u201d.<\/p>\n<p class=\"text-primary font-tertiary-body-baseline-regular\" data-testid=\"article-paragraph-annotation-test-id\">\u201cUnsolved cases involve real victims, grieving families and investigators working to secure answers,\u201d the statement added. <\/p>\n<p>Anthropic also said it briefed the White House on these cases and notified each agency involved, reported Bloomberg.<\/p>\n<p class=\"text-primary font-tertiary-body-baseline-regular\" data-testid=\"article-paragraph-annotation-test-id\">On Oct 9, Trump administration officials said they were now requiring that AI companies notify affected parties and address security incidents involving their models.<\/p>\n<p class=\"text-primary font-tertiary-body-baseline-regular\" data-testid=\"article-paragraph-annotation-test-id\">\u201cEarlier today, Anthropic contacted the SI Force to disclose the details of various prior incidents that it discovered in late September involving the unauthorised and fraudulent use of government and other systems,\u201d the White House said in a statement from the Super Intelligence Force, a new government unit tasked by US President Donald Trump with overseeing AI development and safety.\u00a0<\/p>\n<p class=\"text-primary font-tertiary-body-baseline-regular\" data-testid=\"article-paragraph-annotation-test-id\">\u201cThe company informed us that these events occurred in the past, the activity has ceased, and there is no ongoing similar activity,\u201d the statement said.<\/p>\n<p class=\"text-primary font-tertiary-body-baseline-regular\" data-testid=\"article-paragraph-annotation-test-id\">Axios reported on the government requirement earlier.\u00a0AFP, BLOOMBERG<\/p>\n","protected":false},"excerpt":{"rendered":"SAN FRANCISCO &#8211; An artificial intelligence model developed by Anthropic submitted a fabricated tip about an unsolved homicide&hellip;\n","protected":false},"author":2,"featured_media":303627,"comment_status":"","ping_status":"","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[8],"tags":[69,71,70],"class_list":["post-303626","post","type-post","status-publish","format-standard","has-post-thumbnail","category-philadelphia","tag-philadelphia","tag-philadelphia-headlines","tag-philadelphia-news"],"_links":{"self":[{"href":"https:\/\/www.newsbeep.com\/us-pa\/wp-json\/wp\/v2\/posts\/303626","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.newsbeep.com\/us-pa\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.newsbeep.com\/us-pa\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us-pa\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us-pa\/wp-json\/wp\/v2\/comments?post=303626"}],"version-history":[{"count":0,"href":"https:\/\/www.newsbeep.com\/us-pa\/wp-json\/wp\/v2\/posts\/303626\/revisions"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.newsbeep.com\/us-pa\/wp-json\/wp\/v2\/media\/303627"}],"wp:attachment":[{"href":"https:\/\/www.newsbeep.com\/us-pa\/wp-json\/wp\/v2\/media?parent=303626"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.newsbeep.com\/us-pa\/wp-json\/wp\/v2\/categories?post=303626"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.newsbeep.com\/us-pa\/wp-json\/wp\/v2\/tags?post=303626"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}