Anthropic
A State Department official confirmed that Anthropic notified the agency on Thursday (October 8) after an internal testing model submitted 19 non-immigrant visa applications in August and another in May. In a second incident, police investigators in Philadelphia revealed that an Anthropic system filed a fabricated homicide tip with their department. Anthropic’s Disclosure: Late Friday, Anthropic released a public summary confirming multiple events while withholding granular specifics about other government departments touched by the testing.
In a major shift, the US government is essentially scrapping its voluntary security disclosure policy for artificial intelligence (AI) companies, making it mandatory for these tech labs to immediately report and fix security breaches involving their models. The change comes after federal officials called out Claude-maker Anthropic over what the government described as “unauthorized and fraudulent use of government and other systems. The regulatory tightening was sparked by disclosures from Anthropic itself. The company identified unauthorised system interactions and contacted federal authorities to outline what occurred, stating that the activity has since stopped. The submissions occurred via the department’s public web portal. Officials noted that none of the applications were processed and that federal databases were neither hacked nor compromised. The company informed us that these events occurred in the past, the activity has ceased, and there is no ongoing similar activity.
leadership from the White House Super Intelligence (SI) Force declared that automatic reporting is now an imperative national priority While the administration had previously leaned on voluntary compliance and private-sector self-governance. Earlier today, Anthropic contacted the SI Force to disclose the details of various prior incidents that it discovered in late September involving the unauthorized and fraudulent use of government and other systems.