Language models are starting to author original academic research. A Tokyo-based firm’s AI-written submission performed well at a 2025 workshop, while Google’s Gemini model recently produced a paper on arithmetic geometry that gave novel answers to open questions. In February, the Journal of AI-Generated Papers became the first to exclusively host such material, where AIs are listed as authors and humans as prompters.

A disturbing new AI scam is targeting the owners of lost pets. In recent weeks, dog owners in both Florida and California received fake, AI-generated videos of their pets in danger or in the hospital, along with demands for thousands of dollars for the pets’ “release.” The videos were not real; scammers are casting nets on Facebook for any missing-animal groups that include photos, which can then be recreated into videos with AI. At least one U.S. senator called on the federal government to issue an alert and increase penalties.

A super-powered digital agent could soon be helping police fight financial crimes. (The Wall Street Journal)

San Diego police dispatchers are using AI to answer non-emergency calls and route calls by type. (San Diego Sheriff)

Eight U.S. companies were cleared for up to $1.6 billion in contracts to operate AI on Pentagon networks. (The Hill)

A running database of court filings is tracking over 1,300 cases involving hallucinations or fake citations across the globe. (Damien Charlotin)