Daily at 8AM KST · Summaries and takeaways from 10 AI articles
We cross-check industry press like TechCrunch and The Decoder, official announcements from OpenAI and Google DeepMind, and community signal from Hacker News.The point isn't what happened but why it matters, tied into one read on the day.
119 days running
1163 stories · 318 terms explained · since 2026-07-23
·what we picked →
2026-11-18
· ~9 min read
· Past edition
10 picked from 95 candidates · ordered by significance
Today's Insight
Nvidia's board ruled out Huang's return on a record quarter, but the SEC and Congress are still asking questions
Nvidia board chair Harvey Jones opened Tuesday's earnings call by ruling out Jensen Huang's return as CEO. President Trump and Elon Musk had pushed for it, and the board did not move. The same day, Huang issued a statement saying he would be honored to return.
Continue reading (3)Show less
The results were strong. Nvidia's third-quarter revenue was $115.8 billion, up 103%, and it expects $130 billion in the fourth. But according to the Wall Street Journal, the Securities and Exchange Commission (SEC) is investigating whether Nvidia's disclosures misled investors, and the stock has fallen more than 10% since Huang left. In Congress, what the Trump administration knew and when has become the question.
Anthropic's IPO filings showed that Nvidia and Google each agreed to buy $10 billion of . After the IPO Amazon will hold a 14.7% stake plus a commitment of more than $100 billion in AWS spending over ten years. Some of Anthropic's biggest investors, like Amazon, also sell it cloud services.
Google made Tasks, in which Gemini makes calls and even pays, free, and added a separate switch that turns off AI training while keeping chat history. The same week, Google's security team reported finding an AI tool on an attacker's server that automates patch analysis and attack code. Tasks pays only within a spending cap the user sets and pauses for confirmation before a purchase.
Signal to watch
Watch whether the SEC confirms its Nvidia investigation and whether Huang answers Congress's questions.
01Wired AI
Press
Covered by 5 more outlets
·2026-11-17·~50s read
Business & Funding
Nvidia board chair Harvey Jones said at the start of Tuesday's third-quarter earnings call that the board does not intend to reinstate Jensen Huang as CEO. Jones said that on November 12 the board announced Huang had voluntarily resigned after disclosure of personal behavior inconsistent with the company's code of conduct, and that the board unanimously accepted the resignation. President Trump, Elon Musk and other Silicon Valley executives have pushed for his return, and the board has not moved. Colette Kress, Nvidia's longtime chief financial officer, is interim CEO.
Why it matters
Why It Matters
The board ruled out a return on its earnings call, the most widely heard moment of the quarter, after a president and billionaire executives had pushed for one, and it took no questions about how Huang left. Record results were announced the same day, yet the story was the CEO's chair, not the numbers. For now the company is run by its chief financial officer.
Google made Tasks in the Gemini app free starting November 17. Switching the home-screen toggle from Chat to Tasks lets Gemini connect to email, calendar and payment methods and handle multi-step jobs such as booking appointments, checking bills and grocery shopping. It is available in the US, Japan, Korea, Canada, Australia and India, while the advanced features, such as calling businesses and paying with Google Pay, roll out in the US first. According to The Verge, Tasks is the new name for Spark, which was open only to Google AI Pro and Ultra subscribers when it launched in May.
Why it matters
Why It Matters
At checkout Gemini uses tokens from Mastercard and Visa, stays within a spending cap the user sets, and pauses for confirmation before a purchase. Each task runs in an isolated virtual machine dedicated to the user's account and never shared with other users. Google launched this agent for paying subscribers and is now giving it to everyone for free. Some sites still require a person to click the final button, and Google says it has not eliminated every mistake.
What to do now
Flip the toggle on the Gemini home screen from Chat to Tasks to try it. Set a spending cap before you let it pay for anything.
03TechCrunch AI
Press
Covered by 3 more outlets
·2026-11-17·~40s read
Business & Funding
According to filings Anthropic disclosed on November 17, Amazon will own 14.7% of Anthropic after its IPO and Google 13.7%, together about 28.4%. Amazon converted part of the notes it received for its 2023 and 2024 investments into , and converting all of them would lift its stake to 23.8%. On top of converted notes, Google agreed to buy $10 billion of stock in the IPO's . Anthropic has also committed to spend more than $100 billion on AWS over the next decade.
Why it matters
Why It Matters
Anthropic's two biggest shareholders are also its biggest suppliers. On top of its stake, Amazon has a commitment of more than $100 billion in AWS revenue over ten years. Both companies hold non-voting shares, so their ownership share does not translate into the same share of votes.
The Wall Street Journal reports that the Securities and Exchange Commission is opening a civil investigation into whether Nvidia's disclosures misled investors. It follows Jensen Huang's resignation as CEO after reports of extramarital affairs, one of them with Alice Huang, a Megaspeed executive accused of being a Chinese spy. The Commerce Department has investigated Megaspeed since late 2024 over suspicion that it diverted some Nvidia chips to China, and Nvidia's stock has fallen more than 10% since Huang left.
Why it matters
Why It Matters
The SEC's question is what Nvidia told investors, not the affair itself. Alongside the Commerce Department's Megaspeed probe, Congress has demanded that Huang answer questions, and Treasury Secretary Scott Bessent has publicly criticized Nvidia, so the China chip issue now reaches regulators, investigators and politicians at once.
Nvidia's third-quarter revenue was $115.8 billion, up 103% from a year earlier and above the $108 billion the company had expected. Data center revenue was $107.6 billion, up 110%, and edge computing, which covers gaming PCs and laptops, was $8.2 billion, up 42%. Profit hit a record $65.3 billion, and Nvidia expects fourth-quarter revenue of $130 billion.
Why it matters
Why It Matters
Interim CEO Colette Kress led with results, saying Vera Rubin, Nvidia's next AI chip platform, is off to an explosive start, while board chair Harvey Jones said the call would not address the circumstances of Huang's resignation or the ongoing investigation. The call covered the numbers and left the leadership question unanswered.
In a statement on Tuesday, Jensen Huang said he would be honored to return to Nvidia, and President Trump is backing the campaign to bring him back. A day earlier, the nonprofit Tech Oversight Project, which campaigns to rein in Big Tech, sent billboard trucks around Washington, DC urging Congress to investigate the Trump administration's ties to Huang. CNN reported that Trump may have learned as early as this spring that Huang could be a target of Chinese espionage. Even so, Trump seated Huang at the head table at a state dinner with China's President Xi Jinping and awarded him the National Medal of Science last month.
Why it matters
Why It Matters
The issue has moved from Huang personally to what the Trump administration knew and when, and to the lobbying for looser chip exports. According to the Wall Street Journal, Sen. Richard Blumenthal asked the White House to preserve all communications between Trump and Huang, and Democrats who take over key committees next year have signaled investigations. So the issue will remain even after Nvidia picks a new CEO.
Google is splitting chat history and AI training into separate toggles in the Gemini app's activity settings. Before, the only way to keep chats out of training was to turn history off, which made it impossible to return to a conversation later. Opting out of training is not the default, so users must find the toggle under Activity in Gemini settings and switch it themselves. The change applies only to future conversations, and data already used for training is not deleted. It is rolling out globally and may take a few weeks to reach everyone.
Why it matters
Why It Matters
People who turn history off end up with longer retention: Google will keep chats with history disabled for two months instead of 72 hours. Even with training turned off, Google may use the data in other ways for up to three years.
What to do now
Find the training toggle under Activity in Gemini settings and switch it off. The same menu is reachable from your Google account under Data and Privacy > My Activity > Other activity > Gemini Apps Activity.
08Wired AI
Press
Covered by 2 more outlets
·2026-11-17·~50s read
Products & Services
Live Rewind, a new Apple Watch feature, shows a transcript of the last 15 seconds of a conversation when you double-tap the digital crown and ask Siri. Siri Recap, launched alongside it, summarizes the conversations it hears over a day, if the user allows it, and deletes the recap after seven days unless saved. Both are opt-in and need an Apple Watch Series 12 or Ultra 4 plus an iPhone 16 or newer running iOS 27. They are unavailable to users under 18 and work only in English; they are in public beta in watchOS 27.3 and expected to launch officially before year's end.
Why it matters
Why It Matters
After a week of use, the Wired reviewer found the transcript turns to gibberish in noisy rooms and that showing a screen mid-conversation feels rude. The nonprofit Fight for the Future points out that the wearer opts in but the people nearby never do. Apple tries to ease this with a small chime when the feature starts and by processing Live Rewind audio on the device so it cannot be accessed.
In a report published November 16, Google's Threat Intelligence Group (GTIG) and Mandiant said they found an AI tool called exploit_agent on a real attacker's server. It chains several large language models to find weaknesses by analyzing published security patches and to carry out attacks without a person in the loop, and the server held traces of targeting French media and education websites. AI doing this work is not new, though. According to AI Times, Anthropic reported in June research that Mythos Preview, in a controlled environment, analyzed security patches and produced working attack code.
Why it matters
Why It Matters
The tool targets vulnerabilities whose patches are published but not yet applied. If AI shortens the time attackers need to analyze a patch and build tooling, the window defenders have to apply it shrinks too. Google and Mandiant also noted that automated attacks leave many traces and are easier to detect, and that AI-written attack code often does not work. Google did not say whether this tool actually broke into anything.
What to do now
Keep automatic updates on for your apps and operating system, and apply security patches as soon as they are offered.
오픈AI, 아동 성착취물 탐지 AI 공개...챗GPT 경고·24시간 이용 정지 기능도 도입
Summary
On November 16, OpenAI released a new image for detecting child sexual abuse material () and said it will add warnings and suspensions to ChatGPT. The classifier flags whether an image is AI-generated and assigns risk scores so reviewers can look at the most likely images first. In early evaluation it found a similar amount of harmful imagery as existing systems while cutting the number of images to review by 48%. A ChatGPT user who requests sexualized content involving children sees a warning, has the conversation deleted and is locked out for 24 hours, and repeat violations can lead to account closure and a report to authorities. The classifier is free through the API for verified child-protection groups and related companies.
Why it matters
Why It Matters
Based on early evaluation, the tool cuts the images a person must review by nearly half. In partner testing, 72% of new images flagged with high confidence were confirmed as real cases, so the rest still need human judgment. OpenAI also drew up guidance with Anthropic and Google for how AI systems respond when a report of child sexual exploitation comes in.
Read this far? We'll send tomorrow's too
The same briefing you just read, every day. A weekly synthesis on Sundays and a monthly report at the start of each month come with it. Sent at 8AM KST, which is the evening before in the US.
Free · no ads · one-click unsubscribe
We use your email only to send the briefing, stored and sent via services in Japan and the US. Privacy policy
Check your inbox
Find this subject in your inbox and press the link to start your subscription.