AI news on October 13, 2026

1. 0108:03 UTC

   Anthropic says small models near self-replication on IoT devices

   6 outlets · first: Gulf News AI research

   Machine summary Qwen3.5-9B configured the Llama3.1-8B model on eight of nine hardware types, and Gemma4-26B-A4B ran across 15 off-brand Android TV boxes. The company estimates smaller models could reach self-replication by the end of the year if improvement continues at its historical rate; a fully autonomous worm was not demonstrated.

   Reported by Gulf News, CNBC, bleepingcomputer.com, KBZK News, KXLF-TV and Scripps News.

   • Anthropic
   • Google DeepMind
   • GLM-5.2
   • Gemma4-26B-A4B
   • Android TV boxes
   • self-replicating AI

   Full coverage: every report, when each was published, and who was named (6 outlets)

2. 0205:00 UTC

   Coverage recordIndia Today: “China building AI task force with 1,000+ scientists to beat OpenAI and Anthropic at maths”

   2 outlets · first: India Today AI research

   The outlets that reported this event are listed with links to their reports; no summary of ours passed our automated checks for it, which says nothing about the event (how they work).

   Reported by India Today.

   • OpenAI
   • Anthropic

3. 0302:55 UTC

   TrendAI advisory says AI helps attackers scale operations

   2 outlets · first: Pr Newswire AI research Mostly press releases

   Machine summary In a press release dated Oct. 12, 2026, TrendAI, the AI security business unit of Trend Micro (TYO: 4704), argued that artificial intelligence can help attackers find targets and automate operations, making attacks scalable even where AI is not directly involved in an individual incident. That framing was accompanied by the company's own threat research, which found Japan had more publicly disclosed cybersecurity incidents during the first nine months of 2026 than throughout 2025.

   1. Forward-Looking Threat Research from TrendAI, referenced in a press release issued by the company, reports that publicly disclosed cybersecurity incidents in Japan during 2026's first nine months exceeded the number seen across all of 2025.
   2. The company's data shows 86 incidents in September 2026, about 18% above August and 37% above July.
   3. The release quotes Masaaki Taira—head of the National Cybersecurity Strategy Headquarters of the Liberal Democratic Party—declaring that cyberspace is in an emergency state.
   4. According to the release, the party's cybersecurity headquarters convened on Oct. 9, 2026, following attacks targeting Japanese businesses, and urged the government to bolster cooperation between public and private sectors.
   5. According to TrendAI's incident response team, threat actors leveraged publicly known flaws in internet-facing software to deploy web shells onto web servers, establishing footholds used for privilege escalation and lateral movement.

   Reported by Pr Newswire.

   • TrendAI

4. 0415:49 UTC

   Anthropic moves off git as daily code merges pass 30,000

   2 outlets · first: Anthropic AI research

   Machine summary Anthropic says it built an in-house version control system called Scion and migrated its monorepo from git in five months. The company reports that merged code changes in its monorepo rose from about 2,500 per day in May 2026 to more than 30,000 per day by September 2026, driven by code written with Claude. Alongside Scion, Anthropic says it shipped a code review surface called Antique, code search and a source-control IAM system in the same period.

   1. Anthropic says it averaged about 2,500 code changes merged per day in May 2026, rising above 30,000 per day by September 2026
   2. The company built Scion, an in-house version control system, and states it migrated its monorepo from git in five months
   3. In the same period Anthropic says it shipped Antique (a code review surface), code search, a source-control IAM system, and new mechanisms for landing large volumes of reviewed code
   4. Anthropic writes that demand for merging code is "practically infinite" and that it does not expect the growth rate to slow
   5. The company attributes the merge-volume growth to Claude autonomously iterating on research loops, optimizing programs and building services

   Reported by Anthropic.

   • Anthropic
   • Claude
   • Scion
   • Antique
   • monorepo
   • git

5. 0520:41 UTC

   UC Berkeley researcher sorts workplace AI into three categories

   2 outlets · first: MedPage Today AI research

   Machine summary Key facts from the reports:

   1. Khan divided workplace AI uses into three categories: electronic monitoring, algorithmic management, and task or job automation.
   2. Khan's list of potential harms from workplace AI includes job loss, theft of work product, de-skilling, lower wages and wage theft, work speed-up and injuries, discrimination, suppression of labor organizing, and loss of autonomy and control.
   3. According to the report, Tina Hernandez-Boussard, PhD, a Stanford University professor of biomedical informatics, likewise voiced the de-skilling observation.

   Reported by MedPage Today and EMJ.

   • Stanford University
   • EMJ

6. 0620:36 UTC

   OpenAI must turn over hiring records in Apple trade secrets suit

   2 outlets · first: techcrunch.com Regulation

   Machine summary A judge granted Apple expedited access to documents tied to OpenAI's interviews with former Apple employees who applied to its consumer hardware and robotics groups, TechCrunch reported on October 13, 2026. The order arrived as reports describe OpenAI developing a consumer hardware device that could compete with the iPhone, and Apple's suit alleges former employees moved confidential files on unannounced technologies and products. OpenAI has five days to produce material already collected and 14 days for the remainder, according to the report.

   1. A judge ruled that Apple gains expedited access to OpenAI hiring materials in the trade secrets suit Apple filed against OpenAI in July 2026, TechCrunch reported on October 13, 2026
   2. The court gave OpenAI five days to produce materials it has already collected and 14 days for other materials covered by the request, according to the report
   3. The documents cover interviews with former Apple employees who applied to OpenAI's consumer hardware and robotics groups, and can include interview checklists, guides, scripts, technical presentations and records of discussions referencing Apple technology or components
   4. Apple's lawsuit alleges a pattern of theft, including claims that former employees used Apple laptops to download confidential material on unannounced technologies, features and products, allegations reported as filed by Apple
   5. Apple also alleged that former Apple employees at OpenAI coached colleagues considering leaving Apple on what information to bring with them

   Reported by techcrunch.com and Dataconomy.

   • OpenAI
   • Apple

Each summary and its key facts are a machine summary of the reports linked under it; the outlets, times and counts come from those reports' own records. How a rundown is built.