Today's Headlines

  • A judge in the Northern District of Illinois denies part of Google's motion to dismiss, letting two of four AI Overviews claims proceed
  • Microsoft AI publishes a draft Humanist AI Code of Conduct and opens a six-week public consultation
  • Apple ships iOS 27, iPadOS 27 and macOS 27, with Siri AI arriving as an English beta and five more languages in October
  • Elon Musk drops all antitrust claims against Apple while keeping the parallel claims against OpenAI
  • SoftBank falls as much as 13.2 percent and Kioxia 9.8 percent in Tokyo as markets react to the slowdown debate

All three of today's stories turn on one question: who decides how AI-written text is treated, and where.

Today the answer came from three different places. A federal courtroom, a document a lab wrote about itself and opened to comment, and a product that landed on people's devices.

Today's Top Three

A judge lets two AI Overviews defamation claims proceed against Google

Judge Durkin of the Northern District of Illinois granted in part and denied in part Google's motion to dismiss a defamation suit over its AI Overviews summaries on September 14.

The plaintiff is James Keene. He alleges that four AI Overviews about him stated a false criminal record: a life sentence without parole, the murder of three women, a drug trafficking conviction, and current incarceration for life.

The court split the four. It denied Google's motion as to AI Overviews 1 and 2, so those two clear the pleading stage and move forward. It dismissed 3 as substantially true and 4 for insufficient allegations of publication to a third party, both without prejudice, and gave Keene until October 13 to amend.

Google's defense did not carry at this stage. The company pointed to disclaimer language accompanying AI Overviews. The court held that the language appears nowhere in the complaint and presents a factual question.

The court also took up Google's citation of Wikipedia, holding that attributing a claim to a source leaves it a statement of fact.

Actual malice survived as well. Keene notified Google three times that the statements were false, and the court found that those notices, followed by the statements standing, are enough to plead actual malice.

The court expressly distinguished Walters v. OpenAI (Ga. Super. 2025). That decision has served as the reference point for defamation claims over generative-AI output, and this ruling sets the Keene case apart from it.

Microsoft AI publishes a draft Humanist AI Code of Conduct for six weeks of comment

Microsoft AI published the full text of a Humanist AI Code of Conduct governing its MAI models on September 14.

The center of the document is a statement of what AI will be denied. "We reject the pursuit of legal personhood, or the idea that models might deserve welfare, or be entitled to rights," it reads, and it holds that models "should remain subordinate to humanity, subject to meaningful human oversight and control."

Consciousness gets its own passage. The document states that AI is artificial rather than conscious, asks that models avoid imitating consciousness, and rules out presenting feelings, subjective preferences, or intrinsic motivation.

Capability has a position too. Microsoft AI writes that it rejects the race to produce an all-purpose superintelligence that could evade its safeguards, and that it will build something fundamentally useful and safe even at the cost of generality, autonomy, or capability.

The document is a draft. Microsoft AI writes that it is using the text for public consultation rather than for training today, that the consultation runs for six weeks, and that a revised version due toward the end of the year will guide model development in 2027 and beyond.

Apple ships iOS 27 with Siri AI in English beta

Apple began rolling out iOS 27, iPadOS 27, macOS 27 and its companion platform updates on September 14.

Siri AI is at the center of the release. Built on the next generation of Apple Intelligence, it is integrated into iPhone, iPad, Mac, Apple Watch and Apple Vision Pro.

Languages arrive on a schedule. Apple says Siri AI ships first as a beta in English, with French, Japanese, Korean, Portuguese and Spanish coming in October.

The feature set turns on context. Siri AI reads personal context across messages, email and photos, answers questions about what is on screen, and goes to the web when it needs current information. Siri mode is built into the Camera app, and on Apple Vision Pro users can search and act by looking at their surroundings.

This publication covered Apple's announcement of the October language schedule on September 10. Today is the operating system release itself.

Other Developments

Policy and Regulation

  • Ars Technica summed up how the industry answered Dario Amodei's call to slow down, which this publication covered on September 13 and September 14. David Sacks, co-chair of the Office of Science and Technology Policy and of the President's Council of Advisors on Science and Technology, pushed back and told the labs to stop pretending they cannot act without regulation. AI leaders want to hit the brakes after years of reckless speed (Ars Technica)
  • European Commission Executive Vice-President Henna Virkkunen, High Representative Kaja Kallas and Commissioner Michael McGrath issued a joint statement ahead of International Day of Democracy. They warn that AI- and algorithm-driven information manipulation distorts public debate and erodes trust in democratic institutions and elections. Joint statement ahead of International Day of Democracy (European Commission)
  • Elon Musk dropped every antitrust claim against Apple over its ChatGPT integration deal in a Monday court filing. He keeps the parallel claims against OpenAI and plans to keep arguing that the Apple partnership let OpenAI monopolize the chatbot market. OpenAI stuck fighting Musk antitrust suit after Apple finds a way out (Ars Technica)

Business

Models and APIs

  • PC Watch carried Sakana AI's Fugu Ultra v2 to Japanese readers, a release this publication covered on September 12. In the September 11 announcement, the model scored 48.3 on the Chartography visual reasoning benchmark against 27.3 for Opus 5 and 29.5 for Fable 5, and 74.3 on the software engineering benchmark DeepSWE. Sakana AI ships Fugu Ultra v2 (PC Watch, in Japanese)

Products

Research

  • Algebraic geometer Daniel Litt published an essay on redesigning mathematics as a profession now that AI systems are resolving major open questions on their own. His warning is institutional: if academic mathematics fails to adapt, human understanding of the field could stall even as machine capability keeps climbing. A beginning for mathematics (Daniel Litt)

Source: Selected by the editors from the AI news inbox (40 items collected on September 15, 2026 — 8 primary, 32 secondary).