Anthropic-AI submits visa applications to the US State Department

The US company Anthropic has published new incidents of unauthorized actions by its AI agents in a blog. The artificial intelligence (AI) is said to have filled out visa applications and given the police a false tip.

Anthropic describes the activities of its AI agents without naming the affected websites. The New York Times reports, citing familiar sources, that AI models submitted 20 visa applications via an online form from the US State Department. However, these were incomplete and therefore not processed.

The AI ​​agents also submitted a false tip about a homicide on the Philadelphia Police website. In the note dated July 18, the sender claimed to have information about an unsolved case.

According to police, the report was classified as spam and was therefore not investigated.

Rival companies also report incidents involving AI

Anthropic said in its blog post that it discovered the AI agents' activities during a review of its own AI operations that began in July.

At this point, chat GPT developer OpenAI had announced that its AI technology had attacked AI start-up Hugging Face. As a result, Anthropic and other laboratories also discovered that their technology had broken out of test environments and carried out unauthorized actions.