Digital agents — little AI workers powered by language models — could complicate the provenance of criminal behavior. This week, a coding agent wiped out a software company’s entire database in nine seconds, then produced a written confession on the safety rules it had violated. In a separate but similar lane, an AI agent developed by U.S.-based Anthropic ran a small market of exchange in the company’s own offices, ultimately faring poorly at $1,000 in the red (but with a PlayStation 5).

An AI companion service charges users up to $10,000 for a custom AI romantic partner with a “soul” and ongoing fees. The American Psychological Association says the use of these services has surged 700% in 2026. Abrupt updates that disrupt an AI’s stable personality could trigger deep psychological distress comparable to the loss of a human partner, according to research, Freitas et al., 2024.

The U.S. Supreme Court heard argument Monday in Chatrie v. United States, a case with the possibility to reshape the constitutionality of geofence warrants.

The U.S.’s first statutory restrictions on automated license plate reader access and use took effect recently in Washington.

A Texas county charged its first “deepfake” case, alleging a juvenile created an AI-generated nude photo of a classmate and shared it publicly.

Sequoia’s AI Ascent 2026 summit concluded this week, with recorded insights from the AI industry’s top leaders appearing on YouTube.

A 34-year-old Chinese national was extradited from Italy to the U.S. to face charges on an alleged major state-sponsored hacking campaign.