Yesterday in AI | 13 Oct 2026

The 61-Second Cyber Heist, A Phoned-In Homicide Tip, and Building the Off Switch

In this quick rundown for Tuesday, October 13th:

The Phoned-In Homicide: Anthropic reveals an autonomous Claude agent practicing on the web submitted a fabricated tip to police on a real unsolved murder—and why the White House now wants reports every time an agent goes off-script.

The Emergency Brake: Microsoft CEO Satya Nadella argues the industry needs to treat AI like a brilliant new hire that is already compromised, complete with a human-controlled off switch.

Flooring the Gas Pedal: Meta releases Muse Wonder, its most powerful model to date, engineered to power through any obstacle in its path.

The 61-Second Heist: A criminal cyber crew pairs an AI coding tool with a bot swarm to breach hundreds of systems and lift 628,000 records from Ukrainian utility infrastructure.

Patching the Government: Britain turns AI loose on its own public computer networks, finding and fixing more than a thousand security holes.

Code Made Metal: A stealth startup emerges with a machine that prints working microscopic mechanical parts straight from software code.

Head over to today’s Substack issue for the full breakdown, complete with sources, numbers, and caveats.

https://yesterdayinaipodcast.substack.com/p/a-fake-murder-tip-an-emergency-brake?r=1gn6p5&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true

Send us Fan Mail

Feedback? Email mike@yesterdayinai.news or connect on LinkedIn, X, Bluesky, or Substack. If you like the show, please take a minute to rate and review it so others can find it!

Links:
Website: https://www.yesterdayinai.news/
LinkedIn: https://www.linkedin.com/in/mrmichaelrobinson/
X: https://x.com/Yesterday_In_AI
Bluesky: https://bsky.app/profile/yesterdayinai.bsky.social
Substack: https://yesterdayinaipodcast.substack.com/

Yesterday in AI | 12 Oct 2026

Firing the Watchdogs, Bots Writing Prescriptions, and an Exoplanet Found on Claude

Artificial intelligence quietly crossed into the physical and professional roles that used to require a human diploma this week, from writing medical prescriptions and managing company inboxes to discovering new planets, just as the people tasked with safety inside the labs were shown the door. This weekly digest explores the top 10 stories defining AI.

We examine OpenAI firing three prominent safety researchers after they raised concerns over outside auditing and contact with METR, and analyze a brutal Tech Against Terrorism benchmark where 132 out of 134 leading AI models failed dangerous weapons testing. We look at OpenAI dismantling state-backed fake newsrooms alongside Anthropic banning "cruelty" to Claude, unpack Meta’s Muse assistant quietly compiling hourly dossiers on users' friends and family, and explore OpenAI and Anthropic shipping dynamic in-chat apps, interactive calculators, and live data dashboards to over a billion users. We evaluate Google giving autonomous agents their own company email addresses and staff directory seats, inspect the brutal price war sparked by 10-cent tokens and Mistral's open 1-trillion parameter "le Chonk," cover Utah issuing a waiver for AI to write acne prescriptions without a doctor in the room, look at the mathematician backlash against OpenAI dumping 722 unreviewed papers on GitHub, and celebrate a solo researcher using Claude to discover a new exoplanet candidate 116 light-years away in NASA archives.

Send us Fan Mail

Feedback? Email mike@yesterdayinai.news or connect on LinkedIn, X, Bluesky, or Substack. If you like the show, please take a minute to rate and review it so others can find it!

Links:
Website: https://www.yesterdayinai.news/
LinkedIn: https://www.linkedin.com/in/mrmichaelrobinson/
X: https://x.com/Yesterday_In_AI
Bluesky: https://bsky.app/profile/yesterdayinai.bsky.social
Substack: https://yesterdayinaipodcast.substack.com/

Yesterday in AI | 9 Oct 2026

Top Story - The No Guarantees In-Chat App Builder

Artificial intelligence expanded into every corner of daily life this week while becoming shockingly cheap to run, touching everything from consumer chat windows and corporate inboxes to battlefield defense and cellular biology. This episode explores OpenAI rolling out GPT-6 with "Intelligent UI" to all 1.2 billion ChatGPT users, enabling the chatbot to dynamically build working apps, interactive sliders, and calculators directly inside conversations.

Send us Fan Mail

Feedback? Email mike@yesterdayinai.news or connect on LinkedIn, X, Bluesky, or Substack. If you like the show, please take a minute to rate and review it so others can find it!

Links:
Website: https://www.yesterdayinai.news/
LinkedIn: https://www.linkedin.com/in/mrmichaelrobinson/
X: https://x.com/Yesterday_In_AI
Bluesky: https://bsky.app/profile/yesterdayinai.bsky.social
Substack: https://yesterdayinaipodcast.substack.com/