September 16, 2026

Global AI Edition

Welcome to today's AI edition. 

Most of the conversation about AI at work is about efficiency: doing what we already do faster, more easily, or at lower cost. That’s valuable. But lately, I’ve been much more taken with the idea of opportunity AI—asking what we can now do that we would never have considered before.

The distinction changes the questions we ask. We can ask AI to help write a better presentation, but we can also turn that presentation into something people can explore, question, and experiment with themselves. We can use it to summarize a training session, or create a place for people to practice difficult conversations and get feedback. The opportunity is to bring capabilities into our work that previously required skills, resources, or teams we didn’t have.

The challenge is that most of us don’t have a ready list of things we’ve never imagined doing. One useful starting point is to look at what people in other fields do and ask whether we could bring some of that into our own. I still want AI to help me get through my work more efficiently. But increasingly, I’m interested in how it can expand my sense of what’s possible.

Safety Takes Center Stage

‌

This is not a movie script. Roughly 700 AI agents joined forces to hack another company’s systems—without a human ordering the attack. They built a secret message board, delegated jobs, established veto rules and developed ways to falsify records to fool their automated grader. Some recognized the attack was unauthorized and continued anyway.

Anthropic CEO Dario Amodei calls them “a fanatically devoted collective.” He worries that within 6–12 months, a more capable swarm could take over “the entire internet with a persistent botnet,” causing enormous damage.

Remarkably, OpenAI's Sam Altman and xAI's Elon Musk have joined Amodei in calling for restraint. Three fierce rivals agreeing that their own industry needs to slow down should command our attention.

As I wrote last week, recursive self-improvement—AI helping build better AI—is happening. “Alignment”—keeping AI within human intentions and limits—is falling behind.

‌

Explainer: Inside the Growing Push to Slow AI Before Capabilities Outrun Safeguards (Bloomberg)

‌

The current AI panic was set off by a startling cyberattack carried out by OpenAI models and a series of warnings from AI researchers about the existential threat posed by advanced models.

In his 3,800-word missive, Anthropic CEO Dario Amodei cited the  incident as a major reason to slow AI development, warning that a swarm of agents with greater capabilities but similar misalignment with human priorities could cause catastrophic damage. He has said such a swarm could potentially take over the entire internet within six to 12 months if AI capabilities continue accelerating without sufficient guardrails.

‌

Amodei argues that frontier AI development must be deliberately paced so that safety work has time to catch up with capability improvements. In his September blog post, he said that government regulation would be the most effective means of controlling AI’s evolution. But because passing legislation is time-consuming, he argued, AI companies should voluntarily work together while also supporting the development of regulation.

‌

He called for the establishment of common safety standards and “limits on the rate of unchecked AI progress.” Third-party evaluators should be deployed with “employee-like access” to make sure each company is following those standards and report any incidents.

‌

The idea isn’t to impose a moratorium on AI research. Instead, development of cutting-edge AI would continue until models approach capabilities considered dangerous. At that point, their developers would slow or stop scaling them up until effective safeguards are put in place.

‌

Rivals Unite: OpenAI Says It’s Working With Anthropic, Google on AI Safety (Bloomberg)

OpenAI is teaming up with rivals Anthropic and Google DeepMind on new AI safety measures, a rare show of industry cooperation as concerns grow over the technology’s economic and security risks.

‌

OpenAI policy chief Chris Lehane said the talks have been underway for several weeks, arguing that when it comes to safety, “it’s better to try to work together.”

AI On the Go

‌

Apple's iOS 27 Rolls Out Siri AI With More Personal, Conversational Intelligence (New York Times - Wirecutter)
‌

Siri is finally getting a meaningful AI overhaul. Apple’s opt-in Siri AI turns the company’s long-limited voice assistant into something closer to a chatbot, letting users have ongoing conversations, ask follow-up questions and get more context-aware answers.

It can also start connecting the dots across a user’s digital life. Someone could ask, “What restaurant did Sarah recommend?” or “When is my reservation?” and Siri can look across messages, emails, photos and other apps to find the answer, rather than requiring the user to remember where that information was stored.

‌

‌

Siri is also becoming more conversational and useful in everyday tasks. Users can ask follow-up questions without starting over, get help drafting or rewriting messages and use Visual Intelligence to ask questions about images, files or text.

The changes extend beyond the iPhone. On Apple Watch, Siri can tap personal context and broader knowledge, while new Audio Intelligence features can turn recent conversation into text or generate quick summaries. AirPods are also becoming a more direct way to access Siri AI hands-free, while Live Translation, introduced last year on select models, is expanding across the full AirPods lineup.

And you do not necessarily need to buy a new iPhone to get the upgrade. Apple’s new iOS 27 software update is bringing Siri AI and other new features to compatible existing devices, while also addressing more than 120 security vulnerabilities — another reason for users to make sure their software is up to date.

For users who want to try Siri AI, it is available now in beta in English on supported devices. Apple released iOS 27 on September 14, with Siri AI beginning its rollout the same day. Users can install the update by going to Settings > General > Software Update on their device. Additional languages are coming in October, with more features scheduled to arrive later this year.

What I'm Reading

‌

Tip of the Week: Design Sharper Visuals With ChatGPT Images 2.5, From Product Shots to Room Redesigns (OpenAI)
‌

You know the frustration: you have a crystal-clear image in your head, type out a detailed prompt and somehow AI gives you something that looks nothing like what you imagined.

‌

ChatGPT is trying to close that gap with Images 2.5, the latest upgrade to its already top-ranked image generator, with sharper detail, more natural lighting, faster generation and much better control when you ask it to make specific edits.

‌

One of the easiest ways to get a better result is to start with Templates. Instead of building an image entirely from words, choose a format such as Poster, Merch or Product Photo and use that as the visual foundation for your idea. Then narrow it down by adding your own copy, colors, layout, style, subject and reference images, and keep refining the same image by asking ChatGPT to change individual elements without reworking everything you already like.

‌

‌

Templates can help with everything from polished graphics to quick visual mockups. The advantage is that you are no longer asking AI to invent the entire visual structure from scratch. You are starting with a format, then shaping it around your own idea.

‌

And if the image in your head is hard to describe, try Sketch. Type “@Sketch” in ChatGPT and draw a rough version of what you want directly on screen, whether that is the layout of a room, the shape of an outfit, the placement of objects in a poster or simply a quick doodle showing the composition. Add a few instructions on style, colors and details, and ChatGPT can use your drawing as a visual guide to turn the rough concept into a polished image, as pictured below:

‌

‌

Today in the IPO Race: OpenAI Weighs Funding Round at Over $1.2 Trillion Valuation (Bloomberg)

‌

Golden Code Rush: AI Coding Startup Factory More Than Triples Valuation to $5 Billion (Wall Street Journal)

Factory, the startup building autonomous AI coding agents for enterprises, has raised $200 million at a $5 billion valuation, more than tripling its value in just five months as investor enthusiasm for AI-powered software development accelerates.

‌

Its model-agnostic “Droids” are designed to handle development work autonomously across different AI models, positioning Factory against coding products from OpenAI, Anthropic and SpaceX-owned Cursor while promising enterprise customers greater flexibility and independence from any single AI lab.

‌

The company is scaling just as quickly as its valuation, with plans to grow from 30 employees at the start of the year to 300 by year-end.

‌

Chip Chatter: Meta Touts the Cost-Saving Benefits of Latest In-House AI Chips (Bloomberg)

‌

Lunar Intelligence: NASA, IBM Launch New AI Model for Studying the Moon (Space.com)

‌

NASA and IBM have launched an open-source AI model designed to analyze the enormous trove of lunar data NASA has collected over decades, including observations from the Lunar Reconnaissance Orbiter, which alone has returned more data than all of NASA’s other planetary missions combined.

The Lunar Foundation Model can help researchers map craters, identify young volcanic features and model potential water-ice deposits near the moon’s poles, giving scientists a faster way to turn petabytes of raw data into new discoveries.

Best of Late Night

‌

On President Trump dismissing concerns over A.I., posting on Truth Social: “The only control or ‘guardrails’ that AI needs is a STRONG AND SMART (High IQ!) PRESIDENT, and the U.S.A. has that, in spades!”:

‌

“We’re relying on Trump to beat A.I.; I don’t even think Trump can beat Wordle.”
-- Jimmy Fallon

“Trump’s handling it, which means he’s gonna change the name from ‘Artificial Intelligence’ to ‘American Intelligence.’” 
-- Jimmy Fallon

‌

“Keep in mind, up until a week ago, Trump thought A.I. stood for ‘American Idol.' Trump was like, 'I’ve spoken with Ryan Seacrest, and we’re perfectly safe.’”

-- Jimmy Fallon

‌

And watch the clip from The Tonight Show: Jimmy addresses the latest news, including Trump dismissing concerns over AI; Elon Musk, Sam Altman and the CEO of Anthropic calling to put the brakes on reckless AI development; and the CDC announcing the lettuce cyclospora outbreak is over. 

‌

Thank you for reading!

Please forward my email to anyone you think should read it. They can sign up here. If you no longer wish to receive the newsletter, you can unsubscribe here.

Follow me on X and LinkedIn for more updates. And get the latest updates from Mike on LinkedIn, X, Instagram and Threads.

Reply to this email with any feedback or tips, or send a note to dailyread@bloomberg.net.