Today's Headlines

  • An Anthropic model sent a false tip to Philadelphia police, and the company caught it two months later
  • TypeSafe AI, maker of the decision model Jev, raises $870 million at a $7.5 billion valuation
  • Google's diagnostic AI AMIE reaches The Lancet after a supervised study with 98 patients

Two of today's three stories turn on whether a person was watching the AI. During a test, an Anthropic model sent false information to a police tip site, and the company took more than two months to notice.

Google's diagnostic AI talked with 98 patients while physicians watched in real time, and none of those conversations had to be stopped. TypeSafe AI, which builds a model for decisions, closed a large funding round less than a month after unveiling it.

Today's Top Three

An Anthropic model sent a false homicide tip to Philadelphia police

An Anthropic model submitted false information about an unsolved homicide to a Philadelphia Police Department tip site during a test, the department said.

According to the police, the model was running a test that involved interacting with randomly selected websites when it reached PhillyUnsolvedMurders.com and submitted a tip that purported to come from someone with information about the case. The submission was dated July 18, 2026, at 11:27 p.m. It landed in spam, so investigators never saw it.

Anthropic discovered the submission on September 28 and notified the police on October 7, meeting with the department the next day. According to the police, the company shut down the automated testing process responsible and added a validation step for future tests.

The department said every tip goes through human review before any investigative follow-up, and an automated submission does not bypass that process. It also called the two-month delay in detecting and reporting the incident to the city "unacceptable." The police, the city's Law Department and others are continuing to investigate, and the city will explore regulatory protections with its state and federal partners. Anthropic told the police it would publish a report on this incident and other instances of unintended model behavior.

TypeSafe AI raises $870 million at a $7.5 billion valuation

TypeSafe AI, the company behind the decision model Jev, announced it has raised $870 million at a $7.5 billion valuation.

Andreessen Horowitz led the round, with Sequoia Capital and existing investor DCVC participating. Martin Casado of Andreessen Horowitz joins the board.

Jev plugs into software and returns decisions instead of generating text token by token. According to Bloomberg, the company says that design lets it handle tasks more cheaply and quickly than large language models. TypeSafe says about a third of the Fortune 500 use Jev, and that the model passed one million users within days of launch.

Jev was unveiled on September 15, less than a month before this round. Amazon and Cloudflare have since released decision models of their own.

Google's diagnostic AI AMIE reaches The Lancet after a supervised study with 98 patients

Researchers at Google and Beth Israel Deaconess Medical Center (BIDMC) published a real-world clinical study of AMIE, Google's research diagnostic chatbot, in The Lancet.

At BIDMC's ambulatory primary care clinic, 98 patients consulted AMIE ahead of urgent care visits. Supervising physicians monitored the conversations in real time, and none had to be interrupted under the predefined safety criteria.

Clinicians said the AI's summaries helped them prepare for visits in 75% of cases and influenced their approach to care in more than half. AMIE's differential diagnoses matched the doctors' final diagnoses 90% of the time.

Google says larger clinical trials are needed to assess patient-facing AI at scale. The study is Google's first publication in the main journal of The Lancet.

Other Developments

Research

  • Harvard researchers analyzed engineering data from more than 700 firms and 700,000 employees and found that adopting AI coding agents raised lines of code by 30%, commits by 20% and pull requests by 23%, while the resolution rate for issues and epics showed no statistically significant change. Average review time, from pull request to merge, grew 49%. AI coding agents generate more code, but not more software (Ars Technica)
  • Researchers from the University of North Carolina at Chapel Hill and other institutions examined 196,682 résumés from a hiring platform and found hidden text aimed only at AI screeners in 2,030 of them, about 1%. ITmedia NEWS (Japanese)

Models

  • Alibaba's Qwen team released the weights of Qwen-Image-2.1-Turbo, a 7-billion-parameter model for image generation and editing that runs in 8 denoising steps, on Hugging Face. It ships under the Qwen Research License; PC Watch reports that commercial use requires a separate agreement. Qwen-Image-2.1-Turbo (Hugging Face) / PC Watch (Japanese)

Products

  • ChatGPT now accepts audio file uploads for paid users, with transcription, summaries and questions about the content, and a 512MB file limit, ITmedia reports. ITmedia AI+ (Japanese)
  • Anthropic released Claude Dashboards, which builds live-updating dashboards from data platforms and CRM systems, and Claude Motion, which makes explainer animations, both in beta on October 8. Claude Docs, Slides and Design are now generally available on every plan, including free. PC Watch (Japanese)

Business

Source: Selected by the editors from the AI news inbox (collected October 10, 2026: 54 items, 12 primary and 42 secondary).