Nvidia's CEO Scandal, Bachchan's Deepfake Win, and an AI That Hunts Bugs
Today's five: Nvidia's governance crisis dressed up as a scandal, Amitabh Bachchan's four-year fight to own his own face, a Gemini chatbot hiding inside Google Pay, a criminal case that finally treats AI impersonation as a crime, and the AI OpenAI built specifically to hack things, on purpose.
- Long Press · Daily Five10 Nov 202601 / 05
aiJensen Huang's affair isn't Nvidia's real problem
Nvidia's board opened an independent investigation on November 8 after CEO Jensen Huang admitted to an affair with Alice Huang, a former executive at Megaspeed, a Singapore firm US officials have probed since 2024 for allegedly helping China dodge chip export bans. The New York Times reported that American intelligence believes Beijing used her to get close to Huang; nobody has been charged, and Nvidia says it found no sign its chips or secrets leaked. None of that stopped the stock from dropping over 4%. The real story here isn't the affair, it's a $4.5 trillion company whose only check on its founder-CEO is a board that found out from a newspaper.
The questionIs Nvidia CEO Jensen Huang involved in a Chinese spy scandal?
In one lineNo criminal charges have been filed. Nvidia's board opened an independent investigation on November 8, 2026, after CEO Jensen Huang admitted an affair with Alice Huang, a former Megaspeed executive US officials suspect Beijing used to gather information on him. Nvidia says it found no evidence of a security breach.
The problem
Nvidia sells the chips every AI company in the world needs, and Washington has spent years trying to keep the best ones out of China through export bans. A scandal at the top of Nvidia puts both the company's governance and America's chip-control strategy under a microscope at the same time.
What changed
On November 8, 2026, Jensen Huang admitted an affair with Alice Huang of Megaspeed, a Singapore data-centre firm that bought $4.8 billion in Nvidia gear and has been investigated since 2024 for alleged chip smuggling to China. Nvidia's board launched an independent probe into whether he was a foreign-intelligence target; no charges have been filed against either Huang.
The catch
Nothing here is proven. Nvidia's board says it has no reason to believe security or IP was compromised, China's foreign ministry calls the spying claim a smear, and even the New York Times admits it doesn't know what information, if any, reached Beijing.
What it means for you
You don't own Nvidia stock to care about this. Every AI tool you use, from Gemini to the model behind your favourite app, runs partly on Nvidia chips, and a cloud over its CEO is a cloud over the AI supply chain for a while.
Questions people ask
Who is Alice Huang and what is Megaspeed?
Alice Huang, also known as Huang Le, is a Chinese citizen and former senior executive at Megaspeed, a Singapore-based data centre company that bought about $4.8 billion of Nvidia equipment. US officials have investigated Megaspeed since 2024 over suspicions it helped China obtain restricted Nvidia chips.
How did Nvidia's stock react to the scandal?
Nvidia shares fell more than 4% after the November 8 memo, extending a slide that had already cost it the title of world's most valuable company to Apple on October 30. Analysts call it a blow to Huang's reputation, though too early to call it a turning point for the stock.
What happens next in the Nvidia investigation?
A board committee with outside legal counsel is reviewing whether Huang was targeted by foreign intelligence. Republican lawmakers want him before Congress by November 18, and Senator Elizabeth Warren is pushing for tighter chip-export rules. No timeline for the probe's conclusion has been given.
SourcesAl Jazeera ↗TechCrunch ↗
- Long Press · Daily Five10 Nov 202602 / 05
newsIndiaOnly Amitabh Bachchan can fully own his own face
The Delhi High Court permanently barred the commercial use of Amitabh Bachchan's name, image, voice and persona on November 5, deciding Amitabh Bachchan v Rajat Nagi & Ors four years after the first interim order in 2022. The targets ranged from a fan app called 'Amitabh Bachchan Video Call' to a quiz book, posters, and the domain amitabhbachchan.in, and the ban now extends to anyone else doing the same thing under a different name. It's a solid win, and also a reminder of what it actually took to get there: four years, a famous name, and a court appearance in person when talks broke down. Most people whose face gets stolen for a deepfake ad don't have any of that.
The questionDid Amitabh Bachchan win a personality rights case in court?
In one lineYes. On November 5, 2026, the Delhi High Court passed a permanent injunction in Amitabh Bachchan v Rajat Nagi, barring named defendants and unidentified future infringers from commercially using his name, image, voice or persona without consent, covering a fan app, a quiz book, posters and a domain.
The problem
AI tools have made it trivial to clone a face, a voice or a persona and slap it on an app, a poster or a fake product endorsement, and Indian law has had to catch up case by case rather than with one clear statute.
What changed
On November 5, 2026, Justice Purushaindra Kumar Kaurav of the Delhi High Court passed a final decree in Amitabh Bachchan v Rajat Nagi & Ors, permanently restraining named defendants and unidentified future infringers from commercially exploiting his persona, and disposing the four-year-old suit while leaving the door open for fresh action on new violations.
The catch
This protection took four years, a celebrity's legal team, and Bachchan personally appearing in court once settlement talks broke down in October. It says nothing about how fast or cheap this kind of relief is for someone without his resources.
What it means for you
If your photo or voice gets misused online, this case is the playbook, not a guarantee: document everything, go for an injunction that covers unnamed copycats too, and know that courts are now willing to grant it.
Questions people ask
What exactly did the court ban?
The Delhi High Court permanently barred use of Bachchan's name, image, voice and persona in an app called 'Amitabh Bachchan Video Call,' a quiz book carrying his photo, posters, and the domain amitabhbachchan.in, and extended that ban to anyone else doing the same, even if not yet identified.
Can ordinary people get this kind of protection?
Not easily. This case took four years, started with a 2022 interim order, and needed Bachchan to personally show up in court in October 2026 when settlement talks stalled. Most people facing a deepfake or fake endorsement don't have the money or standing for that fight.
What is a 'John Doe' order in a case like this?
It's an injunction that also binds defendants the court hasn't identified yet, useful against anonymous websites, apps and social accounts that keep popping up under new names. Bachchan's decree used this, so future copies of the same scam don't need a fresh lawsuit each time.
SourcesBar and Bench ↗
- Long Press · Daily Five10 Nov 202603 / 05
tipIndiaGoogle Pay quietly got a Gemini-powered money coach
Google Pay quietly added a Gemini-powered chatbot called Ask Google Pay to its India app, and most people still haven't noticed it. Open the app, tap Ask Google Pay, and ask it anything about your money in any of 10 Indian languages: where your spending went this month, how your SIP is actually doing, or whether you qualify for more credit. It's genuinely useful for the first two. The credit piece is Google nudging you toward a product, so use it, but don't treat every suggestion as neutral advice.
The questionWhat is Ask Google Pay and how does it work?
In one lineAsk Google Pay is a Gemini-powered chatbot built into the Google Pay app in India, launched in 2026. Tap the Ask Google Pay option, type or speak a question in any of 10 Indian languages, and it breaks down your spending, explains your SIPs, or checks your credit options.
The problem
Most people use Google Pay only to send and receive UPI payments, and never dig into the spending and investment data the app is already sitting on.
What changed
Ask Google Pay, a Gemini-powered chatbot inside the Google Pay India app, lets you type or speak a question in any of 10 Indian languages and get a straight answer about your spending patterns, your SIP performance, or available credit, pulled from your actual transaction history.
The catch
It's also a sales channel. The same assistant that explains your SIP will just as happily nudge you toward a credit card or loan, so treat the money insights as useful and the credit suggestions as marketing.
What it means for you
Open Google Pay, tap Ask Google Pay, and ask it where your money actually went last month. Takes ten seconds and tells you more than scrolling through a transaction list ever will.
Questions people ask
How do I open Ask Google Pay?
Open the Google Pay app, look for the Ask Google Pay option on the home screen or search bar, and type or speak your question, like 'how much did I spend on food this month' or 'explain my SIP returns.' It replies in the same 10 Indian languages the app already supports.
Is Ask Google Pay the same as a regular chatbot?
No. It's connected to your actual UPI transaction history inside Google Pay, so answers are based on your real spending and investments, not generic advice. It can also walk you toward more credit options, which is useful but worth treating as a sales nudge, not neutral advice.
Does Ask Google Pay work for everyone in India?
It's rolled out broadly across the Google Pay India app as of late 2026, in 10 Indian languages, so most users should see it on an updated app. If you don't, check for an app update first.
SourcesTechlusive ↗
- Long Press · Daily Five10 Nov 202604 / 05
hot takeAI-voice fraud just stopped being just a lawsuit problem
US prosecutors indicted Jacob Soall, Sharon Hazan and Guy Sheetrit for allegedly using AI voice cloning and deepfakes to impersonate CEOs and lawyers in a scheme accused of stealing $129 million, news that broke in early November 2026. This isn't a civil injunction like the celebrity persona-rights cases; it's a federal criminal indictment, and that's the actual shift here. For years, 'CEO fraud' meant a fake urgent email. Now it means a voice on the phone that sounds exactly right, and prosecutors are finally treating the AI part as an aggravating feature of the crime, not an excuse.
The questionWere people charged for using AI to impersonate CEOs?
In one lineYes. US prosecutors indicted Jacob Soall, Sharon Hazan and Guy Sheetrit, accusing them of using AI voice cloning and deepfakes to impersonate company executives and lawyers in a scheme that allegedly stole $129 million, reported in early November 2026. It's a criminal case, not a civil dispute.
The problem
Voice and video cloning tools are now good enough to convincingly impersonate a real executive, and criminals have been quick to use them for classic wire-fraud scams that used to rely on clumsy fake emails.
What changed
US federal prosecutors indicted three people, Jacob Soall, Sharon Hazan and Guy Sheetrit, over an alleged $129 million scheme using AI-cloned voices and likenesses to impersonate CEOs and lawyers, treating the AI impersonation itself as part of the criminal charge.
The catch
These are allegations, not a conviction, and the case is specific to the US. It doesn't create a general legal standard for AI impersonation fraud everywhere, India included.
What it means for you
If you handle payments at work, the real lesson isn't fear, it's process: any urgent money request from a voice or video call, verify it through a separate channel before you act, no matter how real it sounds.
Questions people ask
How did the AI impersonation scam work?
Prosecutors allege the trio used AI tools to clone the voices and likenesses of real CEOs and lawyers, then used those fake identities to authorize fraudulent transfers or convince victims to move money, a more convincing version of the classic 'CEO fraud' wire-transfer scam that's targeted companies for years.
Is this the first criminal case over AI voice cloning fraud?
No, but it's one of the largest by amount, at an alleged $129 million. Civil courts have handled plenty of deepfake and persona-misuse disputes, India's Bachchan case among them; this is squarely a federal criminal indictment, which carries real prison time, not just an injunction.
How can companies protect themselves from AI voice-cloning fraud?
Treat any urgent payment request from a 'CEO' or lawyer over phone or video as suspicious by default, verify through a separate known channel before moving money, and train finance teams specifically on AI-cloned voice and video scams, not just old-school email phishing.
SourcesSCMP ↗
- Long Press · Daily Five10 Nov 202605 / 05
aiOpenAI built an AI that finds security holes for you
OpenAI built a model specifically for cybersecurity work: GPT-6 Cyber, aimed at vulnerability research and exploit development, with tighter safeguards so it's less likely to go after systems outside its assigned scope. It's rolling out to Daybreak Red, OpenAI's enterprise security subscribers, not to the general ChatGPT crowd. That gatekeeping is the actual point. A model this good at finding software holes is exactly the kind of tool you want locked down to people running authorized security tests, not sitting in a free-tier chatbot anyone can prompt.
The questionWhat is OpenAI's GPT-6 Cyber model used for?
In one lineGPT-6 Cyber is OpenAI's dedicated cybersecurity model, built for vulnerability research and exploit development, with tighter safeguards so it's less likely to attack systems outside its intended scope. It's now available to Daybreak Red, OpenAI's enterprise security subscribers, not to the general public.
The problem
AI models have gotten good enough at finding and exploiting software vulnerabilities that the same capability helps both security researchers defending systems and attackers trying to break into them.
What changed
OpenAI released GPT-6 Cyber, a dedicated cybersecurity model built for vulnerability research with improved scope adherence, and is rolling it out to Daybreak Red, its enterprise security subscriber tier, rather than to general ChatGPT users.
The catch
Restricting it to paying enterprise security subscribers reduces but doesn't eliminate misuse risk, and the model's actual effectiveness against real-world threats outside OpenAI's own benchmarks hasn't been independently tested yet.
What it means for you
If you run a security team, this is a tool worth evaluating for authorized penetration testing. If you don't, the real takeaway is simpler: AI-assisted vulnerability research is now mainstream enough that OpenAI built a dedicated product for it.
Questions people ask
Can anyone use GPT-6 Cyber?
Not yet. It's rolled out specifically to Daybreak Red subscribers, OpenAI's enterprise security tier, rather than to ChatGPT's general consumer or developer audience, because a model built to find and exploit software flaws needs tighter access control than a general chatbot.
Why does OpenAI need a separate cybersecurity model at all?
General AI models are already good enough to help both attackers and defenders find software vulnerabilities, so OpenAI built a dedicated version aimed at defenders, with improved 'scope adherence' so it stays inside the systems it's authorized to test instead of wandering into ones it isn't.
Does this mean AI can now hack on its own?
It means AI is now good enough at vulnerability research that companies are building specialised, access-controlled models for it rather than relying only on general-purpose chatbots, which is a meaningful jump from a year ago, though it still works inside defined security engagements, not autonomously.
SourceseSecurity Planet ↗
Taps
Quick hits. One line each, for when a long press is too much.
- IndiaMediaNama ↗
RBI data: credit cards now drive spending, debit cards just withdraw cash
September numbers show credit cards driving spending growth while debit cards are now mainly a cash-withdrawal tool, UPI having taken point-of-sale. MediaNama
- IndiaMediaNama ↗
Meta actioned 20.21 million pieces of India content in September
Instagram enforcement rose while Facebook's fell, and Meta's own proactive-detection rate for harmful content dropped the same month. MediaNama
- GlobalGizmodo ↗
Binance's CFO steps down a week after the $6.3 billion hack
Kaiser Ng is out; the hack, blamed on North Korean hackers, had funds moved through THORChain. Gizmodo
- GlobalFree Press Journal ↗
Anthropic could start its IPO roadshow before Thanksgiving
Marketing could begin as early as November 9, with talk of a valuation up to $2 trillion; investors are reportedly unconvinced. Free Press Journal
- GlobalYicai Global ↗
DeepSeek's Harness agent app hits 1.54 million daily users
The company shared daily-user numbers for Harness for the first time, with 5.5 million monthly users. Yicai Global