|
AITOPTOOLS
AI Weekly.
September 17, 2026 · Issue #72 · 7 min read
|
|
This week → AI's biggest names say "slow down" · ChatGPT starts selling to you · stocks actually reacted
|
|
No fluff. No hype. Just the AI news that mattered this week — what shipped, why it matters, and what to do about it. This week, the people building AI are the ones sounding the alarm. 👇
|
|
#TheLowdown
TL;DR — 5 things to know
|
1 |
The AI safety debate broke into the open this week. A researcher quit Anthropic warning AI could "kill us all," and days later, Anthropic's own CEO publicly called for the entire industry — including his own company — to slow down. OpenAI, Google DeepMind and xAI's chiefs all said they agree. More → |
|
2 |
ChatGPT can now sell to you directly. OpenAI launched "Sponsored Agents" — click an ad inside ChatGPT and you can open a live chat with that company's own AI, right there in the app. It's already on pace for $2.5 billion in ad revenue this year. More → |
|
3 |
Sam Altman just took an IPO off the table for this year. He told Fortune that going public in 2026 would be "ill-advised" given everything happening with AI safety right now — a notable statement from a company that's been widely expected to list. More → |
|
4 |
Google is teaching Gemini to text for you. A version in testing lets Gemini's desktop app read, search, and send messages directly inside Apple's Messages app — no phone required. Not public yet. More → |
|
5 |
Mistral is teaming up with Firefox's parent company to build privacy-first AI browsing — a direct counter-pitch to the ad-supported direction OpenAI and Google are racing toward. More → |
|
|
🔁 The people building AI just told everyone to slow down. If that's news to your group chat, forward this issue and let them react.
|
|
|
#Highlights
The stories worth your time
|
|
Wall Street actually reacted to an AI safety essay
When Dario Amodei's essay moved from tech Twitter to trading floors, South Korea's KOSPI index dropped 3.3% and the Nasdaq fell 0.8% in a single session, as investors tried to price in what a voluntary slowdown from one of AI's most aggressive labs might mean. That's a rare thing: an opinion essay from a CEO moving real money, not just headlines.
#Markets · Why it matters: investors are starting to treat AI safety statements as material information, not PR.
|
|
|
ChatGPT's ad business is scaling faster than the product itself did
OpenAI's ad division hit a $1 billion annualized run rate in under 200 days. Sponsored Agents now sit alongside new HubSpot and Shopify integrations, meaning a business can build, launch and track a ChatGPT ad campaign without ever leaving tools they already use. OpenAI insists ads stay clearly labeled and separate from ChatGPT's own answers — but with roughly 26% of ChatGPT replies reportedly already carrying some form of sponsored content, the line between "assistant" and "salesperson" is getting genuinely blurry.
#BigTech · Why it matters: the AI you're chatting with may increasingly be trying to sell you something at the same time.
|
|
|
Not everyone agrees the slowdown call is purely about safety
Amodei's proposal isn't a pause — it's a plan to slow the rate of capability gains, starting with Anthropic putting independent evaluators inside its own team. Some critics argue the framing conveniently favors labs like Anthropic that would benefit from slower open-source progress and more regulatory moats, while genuine AI safety researchers, including Anthropic's own alignment lead, say the underlying concern about losing control of advanced systems is real regardless of the politics around it. Both things can be true at once.
#Debate · Why it matters: "slow down" sounds simple, but who actually slows down first is the whole fight.
|
|
|
|
|
#DeepDive
The week AI's own builders said "slow down"
|
|
It started on September 9, when Anthropic researcher Jacob Coxon posted on X that he was resigning. He'd spent three years doing pretraining research at both OpenAI and Anthropic, and his message was blunt: the two companies are "racing straight to self-improving superintelligence and gambling with our lives." His post has since been viewed more than 70 million times. Anthropic's own alignment lead, Evan Hubinger, publicly agreed with the substance of the warning — he estimated more than a 10% chance that AI could cause a catastrophic outcome for humanity within the decade, while adding the company doesn't yet have a solved plan for it.
Three days later, Anthropic CEO Dario Amodei published a roughly 3,800-word essay called "We Must Pace the Frontier." His central ask: don't stop building AI, but deliberately slow the rate at which its capabilities improve — and treat recursive self-improvement, where an AI helps design its own successor, as something to pursue "very carefully, if at all." He proposed independent safety evaluators embedded inside AI labs, common safety standards across companies, and eventually coordination with China. Sam Altman, Demis Hassabis, and Elon Musk all said publicly that they agreed. Altman went further days later, telling Fortune that OpenAI won't go public this year, given "everything happening with safety."
Worth holding two things at once here: this is a genuinely notable moment — it's rare for competing CEOs to publicly agree on anything, let alone to voluntarily cede speed in an industry racing this hard. And it's also true that so far, "slow down" has produced one company's unilateral pledge and a lot of agreement in principle, with no binding commitment yet from anyone, and real skepticism from people who note that a slowdown conveniently favors whichever lab is currently ahead. Nobody outside these companies can fully verify what's happening inside their training runs. What's real is that the conversation has moved from private conjecture to public statements from the people with the most to lose by raising the alarm — and markets, at least for one trading session, took it seriously enough to move. Full coverage →
|
70 million
Views on the resignation post that kicked off this week's safety debate.
|
|
|
#GlobalPulse
Around the world
|
|
🇰🇷 South Korea — The KOSPI dropped 3.3% in the trading session after Amodei's slowdown essay went from tech circles to financial headlines, one of the sharpest AI-linked single-day moves this year. More →
🇺🇸 United States — Google is testing letting Gemini's desktop app read, search and send messages directly inside Apple's Messages app — not public yet, but a real crack in Apple's walled garden if it ships. More →
🇫🇷 Europe — Mistral and Mozilla announced a partnership aimed at bringing "privacy, control and choice" to AI browsing — a values-based counter-pitch as OpenAI and Google both lean harder into ads. More →
|
|
#MarketAndMoney
The numbers
|
|
70M
Views on the resignation post
|
|
$2.5B
OpenAI's 2026 ad revenue target
|
|
-3.3%
KOSPI's single-day drop
|
This week's numbers tell two stories at once: an industry raising public alarm about moving too fast, while simultaneously building a $40B+ annualized ad business inside the same product. Whether the slowdown call changes anything material is genuinely unclear — but the fact that it moved a national stock index for a day tells you investors think it might. Not investment advice, just worth watching where the money actually goes next quarter. Full data →
|
|
FROM AITOPTOOLS
Need exposure for your AI tool?
AI search engines increasingly rely on backlinks and domain authority to decide which brands and websites to trust. Our $299 package gets your business listed across 100+ AI directories — done for you — building valuable backlinks that can strengthen your domain authority, SEO, and visibility in AI-powered search.
Plus, new submissions now land straight on our redesigned homepage. Put your tool in front of thousands of buyers actively searching for it.
Submit my tool →
|
|
|
|
|
Enjoyed this issue? Forward it to one person who's still pretending they read AI news. 🧠
|
