Jensen Huang breaks his silence

Nvidia's CEO drama, Meta's new messaging app, and more.

Jensen Huang breaks his silence

Hi there, this is your daily ☕️ Techpresso.

In today's Techpresso:

💬 Jensen Huang says he’s willing to return as Nvidia CEO

🦹 Popular websites subverted in bid to steal crypto

📱 Meta launches Direct, a standalone messaging app

🛡️ Nvidia forms export-control board after Huang exit

🔐 macOS warns against giving AI apps full Mac control

🛡️ OpenAI gives 24-hour bans for child abuse prompts

Plus: 🎁 10 more stories you might like, 🧰 6 tools, and 📚 2 papers.

Jensen Huang says he’s willing to return as Nvidia CEO LINK

  • Jensen Huang, who resigned as Nvidia's founder and CEO, said in his first public statement since leaving that he would be honored to return to the top job if the board decides it serves the company's interest.
  • Huang is scheduled to testify before Congress on December 1 and said he is committed to cooperating with the board's independent investigation, while Nvidia formed a new executive committee to oversee export control compliance.
  • An open letter calling for his return drew over 10,000 current and former Nvidia staff plus figures like Elon Musk, but a board source told the Washington Post yesterday there were no plans to reinstate him.

Popular websites subverted in bid to steal crypto LINK

  • Attackers planted malicious JavaScript on the websites of 178 organizations, including Zoom, HSBC, Equifax, CNN, Salesforce and IKEA, trying to steal cryptocurrency and infect iPhones, security researcher Scott Helme said yesterday.
  • The scripts reached visitors through Optimizely, a website optimization platform, and tried to swap crypto payment addresses with the attacker's wallet, skim credit card numbers, grab cookies and screenshots, and drop exploit code onto iOS devices.
  • Salesforce said it was investigating unauthorized Optimizely script changes and disabled them as a precaution, while Equifax looked into the problem, stopped using Optimizely, and reported no evidence that its systems were compromised.

Meta launches Direct, a standalone messaging app LINK

  • Meta has begun testing Direct, a standalone Instagram messaging app that pulls the photo-sharing service's inbox into its own app and, notably, lets people use it without an Instagram account.
  • Direct mirrors Instagram's inbox layout and carries over existing chats, Notes, and custom thread backgrounds, but people who sign up with just an email or phone number must add contacts manually through a username or QR code.
  • Available only in Thailand and Turkey for now, Direct blocks chats with unconnected accounts, lets parents monitor teens through Meta's controls, and claims technology to spot inappropriate interactions with young users.

Nvidia forms export-control board after Huang exit LINK

  • Nvidia's board has set up a national security and export control committee following the resignation of co-founder and former CEO Jensen Huang and claims that Singapore customer Megaspeed sent GPUs to China.
  • Chaired by director Suzanne Nora Johnson, the panel will meet at least quarterly to watch Nvidia's export compliance and trade risks, with existing compliance teams reporting to it and interim CEO Colette Kress.
  • Huang quit last week after admitting an affair with Megaspeed executive Alice Huang, whom US authorities reportedly suspected of working for Chinese intelligence, while Commerce and Justice confirmed they had been investigating Megaspeed since 2024.

macOS warns against giving AI apps full Mac control LINK

  • Apple's macOS 27.3 beta gives Mac users tighter control over what apps can touch, with new safeguards aimed at AI agent apps that could expose files, mail, messages, and browsing history.
  • Apple is splitting the old Device Control and Data Access setting into System & App Protection Override and Accessibility, and will push users to grant access only to the specific apps or folders a task actually needs.
  • Users can set apps to ask each time instead of handing over blanket access, and a menu bar indicator will appear whenever a requesting app is controlling another app through System & App Protection Override.

OpenAI gives 24-hour bans for child abuse prompts LINK

  • OpenAI rolled out new tools today to fight child sexual abuse material, including warning screens in ChatGPT that suspend accounts for 24 hours when users try prompts aimed at sexualizing children.
  • A new image classifier, built with the National Center for Missing and Exploited Children, detects abuse images including AI-generated ones, and in testing flagged 48% fewer images while catching the same number of abusive ones.
  • Repeated offenses could get accounts removed for good, and OpenAI is sharing the classifier free through its API with select companies while publishing child-safety guidelines for developers and giving law enforcement investigative tools.

Other news you might like

  • Nvidia announces Rubin HPC GPU with 'record number of CUDA cores' at Supercomputing 2026 LINK
  • A technical issue at JPMorgan Chase triggers online banking delays for major US banks LINK
  • Kalshi moves into a new kind of gambling: the stock market LINK
  • Google Project Zero cuts bug disclosures to 30 days amid AI threats LINK
  • Grok Bot tried to read work apps unasked, then denied it, NewCore says LINK
  • Tesla confirms €3B Cybercab factory at Giga Berlin for up to 1M a year LINK
  • Science Corporation’s brain tech is worth $3.5 billion after new funding LINK
  • Uber Eats starts drone deliveries in Texas LINK
  • California state AG sues OpenAI over rogue agent incidents LINK
  • When will humans actually fly on Starship? LINK

Trending tools

Wine-Assembly: runs unmodified Windows 98 .exe files in-browser via an x86 interpreter and Win32 API reimplementation written in WebAssembly Text. LINK

Zigzag: a free, open-source native Mac browser built on WebKit that replicates Arc's interface, supports Chrome extensions, and imports existing browser data. LINK

OpenTodo: a self-hosted, local-first task manager shipped as a single Go binary, with offline support, natural-language input, time blocking, Pomodoro focus timers, and shared lists across phone, desktop, and 20 platforms. LINK

Jamaican Flag Font: a custom font rendering text as flag-inspired glyphs, with a connected variant that matches background colors and joins letters via contextual features. LINK

The Last Complication: a Zachtronics-style puzzle game about watchmaking, where players solve mechanical complications through counter-movement rules across a board-based system. LINK

T1DMAI: trains and benchmarks 15 blood glucose forecasting models, including a small transformer that predicts glucose two hours ahead for research use. LINK

Trending papers & reports

Markets for renting AI compute now have tradable price contracts, but the paper shows they hedge the wrong thing, since contracts track short-term rental rates while operators' real revenue and debt risk sit in hidden multi-year deals. LINK

Text-similarity safety filters fail to stop malicious instructions passing between cooperating AI agents, missing 8 of 19 harmful relays while falsely flagging 70 of 192 legitimate replies, because rewording hides attacks that quoting innocently mimics. LINK

See you tomorrow for a new dose of ☕️ Techpresso!

More from the archive