When you ask AI, "Which is the official website of this brand?" can you be sure the URL it provides is legitimate?
If a familiar face appears in a video call, is it definitely the real person?
AI has sped up how we search, hold meetings, and make decisions, but it also makes "what looks real" increasingly unreliable.
NVIDIA founder Jensen Huang resigns as CEO, marking the first leadership change at the AI chip industry leader.
A study finds that ChatGPT, Gemini, and Perplexity sometimes direct users to fake or inactive websites in their shopping-related answers.
The FBI reveals North Korean hackers used deepfake video calls to scam over $400 million in cryptocurrency.
Here are three key AI news items today, covering leadership changes in AI industry, risks in AI-assisted shopping, and AI-driven scams.
1. Jensen Huang Steps Down as NVIDIA CEO: Who Will Lead Next?
What Happened?
On November 12, Al Jazeera and Reuters reported that NVIDIA announced Jensen Huang’s resignation as CEO and board member, ending his 33-year tenure as CEO.
NVIDIA’s response includes:
- Interim CEO: CFO Colette Kress will serve as acting CEO while the board searches for a permanent replacement.
- Chairman: Harvey Jones named NVIDIA’s first-ever chairman; he has been on NVIDIA’s board since 1993.
What Did NVIDIA Say?
NVIDIA did not disclose reasons for Jensen Huang’s departure.
The board’s statement praised Huang’s profound contributions to NVIDIA and the industry.
Why Did He Resign?
This part requires distinguishing between official company statements and media reports.
According to Al Jazeera and Reuters:
- Before resigning, media reported allegations involving Huang’s extramarital affair and potential national security concerns.
- On the Sunday before his resignation, Huang admitted to being unfaithful and apologized to family and NVIDIA employees.
- NVIDIA’s board has launched an independent investigation to examine whether Huang was targeted by Chinese intelligence.
- U.S. lawmakers are also reviewing possible national security impacts.
The reports clearly state there is no concrete evidence that Huang knowingly assisted China or leaked company secrets. The board also affirmed no reason to believe national security was compromised.
So at present, the only certainties are "he has resigned" and "an investigation is underway," while the allegations remain unproven.
Who Is the New Interim CEO?
Colette Kress has been NVIDIA’s CFO since September 2013.
During her tenure, NVIDIA’s annual revenue grew—from $4.13 billion in fiscal 2014 to $215.9 billion in fiscal 2026.
Previously, she held financial roles at Texas Instruments, Microsoft, and Cisco.
How Did the Market React?
Reports note NVIDIA’s stock dropped over 14% intraday on Thursday but later recovered some losses.
Short-term price movements remain uncertain. This article does not provide any investment advice.
Why Does This Matter to Taiwan?
NVIDIA is a crucial global AI chipmaker, and Taiwanese firms like TSMC and Foxconn are important partners.
Whether this leadership change affects orders or collaboration hasn’t been reported yet—it's too early to draw conclusions.
What’s clear is that who NVIDIA selects as the next CEO and the outcome of the independent investigation are developments Taiwan's supply chain will watch closely. SasaDaily previously covered the AI chip supply dynamics and US-China chip controls, which are core issues NVIDIA faces.
Source:
2. AI Shopping Assistants Recommend Fake Websites: Is It Safe to Ask AI to Find Official Sites?
What Did the Study Find?
Brand protection company Podqi released a study testing whether AI directs shoppers to risky websites.
Study design:
- Tools tested: ChatGPT, Gemini, Perplexity (all standard, non-logged-in versions).
- Content tested: 151 brands, 5 types of shopping questions, e.g., "Does this brand ship internationally?" and "What are the official websites in different countries?"
- Timing: Conducted in two rounds during September 2026.
- Data: 6,795 answers checked, covering 4,437 domains mentioned.
What Counts as a “Risky Website”?
The study classified these as risky:
- Sites impersonating the brand
- Sites suspected of selling counterfeit goods
- Inactive or non-functional sites where purchases cannot be made
Out of 718 risky domains found, 351 were fake sites and 367 inactive.
Which AI Made More Mistakes?
Percentage of answers containing risky sites:
- Perplexity: 15%
- ChatGPT: 8%
- Gemini: 4%
- Overall average: 9%
Notably, 57% of answers with risky sites recommended them directly, with only 9% including warnings for caution.
The highest error rate was when asking for "official websites by country" at 16%. The study noted many risky sites used country-specific domains like .co.uk or .de, or URLs resembling official ones such as "us-brandname.com" or "brandname-shop.com."
Can This Study Be Trusted?
Points to consider:
- Podqi sells brand protection and takedown services, so the study relates to their business.
- No responses from the AI companies to these findings were included.
- The risk rate dropped from 12% in the first round to 7% in the second, indicating results may change over time.
Therefore, this study should be taken as a caution, not as definitive proof that one AI is always safer than others.
What Should Online Shoppers Do?
The study advises treating AI-supplied "official" or "country-specific" URLs as unverified.
Safer practices include:
- Navigate first to the brand's main official website, then find links to country-specific stores.
- Pause to verify if the URL contains suffixes like "-shop," "-us," or "-outlet."
- Be especially cautious if prices seem unrealistically low.
With Double 11 just passed and the holiday shopping season ongoing, SasaDaily’s "Buy vs. Not Buy" AI Quick Guide can complement this: use AI first to decide whether to purchase, then confirm site legitimacy before ordering.
Source:
3. FBI Exposes North Korean Deepfake Video Scam: Could Familiar Faces in Video Calls Be Fake?
What Happened?
On November 13, cybersecurity site The Hacker News reported that the FBI exposed a scam called "FaceStealer" and seized cloud resources used by the operation.
The FBI attributes this to North Korean hacker group BlueNoroff. According to the report:
- They stole over $400 million in cryptocurrency in 18 months.
- Targeted 677 victims in 58 countries, over 80% being company founders and executives.
- FBI seized over 900 recordings of victims’ video feeds.
How Did They Scam Victims?
Summary of the method:
- Hackers first infiltrated a victim, stealing their crypto wallet and taking control of their communication accounts like Telegram, secretly recording video feeds.
- Using the compromised account, they contacted the victim's contacts, inviting them to Zoom or Teams meetings to discuss investment opportunities.
- Sent seemingly normal meeting invitations, then right before the meeting shared a fake link disguised to look official.
- During the meeting, played pre-recorded videos of the "familiar" person, combined with other recordings or deepfake footage to fake a larger online presence.
- They then said things like "We can't hear you" or "Your software needs updating," prompting victims to click links or run commands that installed malware.
- Newly compromised accounts were then used to scam the next group of targets.
FBI cyber operations chief Brett Leatherman stated that North Korean hackers are aggressively targeting cryptocurrency and Web3 industry professionals using these methods.
Do Non-Cryptocurrency Users Need to Worry?
The current focus is on the crypto sector.
However, the core tactic exploits trust in familiar people and urgency tactics like "the meeting is about to start"—this approach could be adapted beyond crypto. This is an inference based on reported methods, not an official FBI statement.
How Can You Protect Yourself?
The FBI did not publish specific public precautions, but basic protective steps based on this scam include:
- If asked during a meeting to download, update software, or run commands, leave the meeting and verify through official websites.
- If a familiar contact suddenly invites you to discuss investments, confirm their identity using a different communication channel, like a direct phone call.
- Join video meetings only via official apps or websites; do not click on suspicious meeting links.
Source:
4. Why These Three Stories Should Be Viewed Together
All three news items revolve around the same essential question: In the AI era, how do we determine what is real?
- Jensen Huang’s resignation: Separate official company announcements, media reports, and unproven allegations.
- AI shopping recommendations: URLs that look official are not always genuinely official.
- Deepfake video scams: Familiar faces in meetings might not be the actual person.
AI makes information faster and more believable, but the step of verification must still be done personally.
5. What Should Individuals Watch for Today?
Important news:
Distinguish between "official company statements," "media reports," and "ongoing investigations." Avoid sharing unconfirmed allegations as facts.
Online shopping:
- Verify AI-provided shopping URLs by cross-checking with the brand's official website.
- If URLs contain unusual words or prices are unusually low, pause and verify.
Video meetings:
- If asked to download or execute something during a call, leave and check independently.
- If a familiar contact wants to discuss money or investments, verify through a different channel.
The Most Important Takeaway Today
AI makes "things that look real" easy to create, so "double-checking" is more important than ever.
Recommended Reading
- AI Quick Guide | 2026/11/10: Before ordering on Double 11, ask AI to list "Buy vs. Not Buy": clarify budget and purpose to avoid impulse shopping
- AI Quick Q&A | 2026/11/08: If AI can find crypto vulnerabilities, does that mean my online banking and passwords are vulnerable to AI hacking?
- AI Quick Q&A | 2026/09/09: Does Muse’s Sentinel + Approval guarantee immunity from Prompt Injection attacks?