This morning’s briefing leads with Nvidia’s leadership fight, followed by South Korea’s exam-cheating scandal and new child-safety research. There are also US policy moves and a measure of the AI investment downturn. It is a quiet morning for workplace tool releases.
Nvidia faces a Huang comeback campaign and US demands for change
What happened. Nvidia’s former chief executive Jensen Huang has discussed returning to the company with confidants, the Financial Times reported on 15 November, according to Reuters. Elon Musk is among more than 165,000 signatories calling for Huang’s reinstatement, The Guardian reports.
US Treasury Secretary Scott Bessent said Nvidia should choose its next chief executive from outside the company. He also called for its board and outside lawyers to answer questions within ten days about compliance with chip-export restrictions, Reuters reports. Huang resigned after revelations of extramarital affairs, including one involving a suspected Chinese spy.
Finance chief Colette Kress became interim chief executive on 12 November and retains her finance role. Nvidia’s announcement says the board’s search for a permanent successor will consider internal and external candidates.
Why it matters. Nvidia’s board says it expects customer commitments, operations and its product roadmap to continue without interruption in its leadership announcement. Our analysis: those commitments are the practical business measures to watch alongside the succession and export-compliance investigations.
The implications extend beyond hardware. Nvidia’s open models give organisations a foundation they can customise with local data and deploy on infrastructure they control. The Financial Times argues that Huang’s departure removes an influential advocate of open AI systems and strengthens Anthropic and OpenAI. For AI buyers, this places the balance between open models and hosted services within the leadership story.
AP reports further falls in Asian AI shares on Monday morning. Pinpoint Asset Management chief economist Zhiwei Zhang said Nvidia’s turmoil had intensified existing concerns about the sustainability of AI capital spending.
US dollars per share
Nvidia’s shares through the leadership crisis
Daily closing share price
People are asking
**Who is running Nvidia?
** Colette Kress is interim chief executive and continues as finance chief, according to Nvidia.
**Has the board brought Huang back?
** Nvidia’s published arrangement puts Kress in charge while it searches for a permanent successor. The comeback discussions are described in Reuters’ report.
**What happens next?
** Nvidia reports quarterly results on Tuesday 17 November, followed by its earnings call at 22:00 GMT. The company confirms that timetable in its announcement.
South Korea confronts AI cheating in professional exams
What happened. South Korea’s top tax accountant examination scorer admitted on 15 November that he cheated in both stages using a miniature camera, smartphone and generative AI. He also admitted using the same method in a certified public accountant examination, Yonhap reports.
The candidate, Han Won-seok, 34, photographed questions and checked AI answers on his phone, according to The Korea Herald. In April’s first stage, he answered 159 of 160 questions correctly; the second stage took place in July, Yonhap reports. His tax-law score was 100 out of 100, against an average of 30.18 among all candidates, The Korea Herald reports.
On 16 November, South Korea’s prime minister announced an interagency emergency response team ahead of the university entrance examination on Thursday 19 November. She also instructed officials to take preventive measures for other state examinations, according to Yonhap’s follow-up.
Why it matters. South Korea’s prime minister said conventional examination management has limits in responding to miniature electronic devices and outside assistance. She called for advanced detection technology and a comprehensive government response, Yonhap reports.
Our assessment takeaway: organisations using tests for recruitment, promotion or certification should state what AI assistance is permitted and assess the candidate’s ability to explain and verify the work. A finished answer and a live explanation provide different evidence of competence. The case makes that distinction concrete: Han described receiving AI answers during supervised examinations, as The Korea Herald documents.
Points out of 100
A perfect score far above the examination average
Tax-law section of South Korea’s first-stage tax accountant examination, 2026
People are asking
**Was he using AI glasses?
** He described a miniature camera and smartphone in his admission, Yonhap reports.
**Did he get only one question wrong across both stages?
** The 159-out-of-160 result belongs to April’s first stage. He admitted cheating in both stages, according to Yonhap.
**What is the practical lesson for workplace assessments?
** Our recommendation is to state when AI assistance is allowed and ask candidates to explain their choices and verify a submitted answer. This follows the examination-security failures described by The Korea Herald.
Childlight reports a surge in AI-related child exploitation
What happened. Childlight, the child-safety institute hosted at the University of Edinburgh, is presenting its 2026 Into the Light findings on 16 November. Its research reports an increase exceeding 8,400% in AI-related child exploitation reports over two years, Press Association reports.
Childlight’s report traces the increase from about 4,700 reports in 2023. The US National Center for Missing & Exploited Children says its CyberTipline received more than 400,000 reports involving generative AI in 2025. Within those submissions, more than 182,000 involved possessing, generating or attempting to generate AI child sexual abuse material, according to its annual data.
The Internet Watch Foundation’s 2026 research report provides a separate measure: it identified 3,443 AI-generated child sexual abuse videos in 2025, compared with 13 in 2024.
Why it matters. The harm reaches identifiable children. NCMEC says offenders manipulate existing abusive imagery to create further content; more than 275 victims of AI-generated abuse material have been identified since it began tracking this in 2023. Its findings describe that repeated victimisation.
Childlight calls for reporting that distinguishes fully fabricated images from altered images of identifiable children and other forms of synthetic material. Its report says these distinctions help investigators identify victims and decide what action to take.
Our practical recommendation for organisations choosing image tools is to examine the provider’s safety testing, misuse controls and reporting process. The IWF’s policy proposals call for mandatory testing before release, protections built into the product and effective content moderation.
Videos
Identified AI-generated abuse videos rose sharply
Videos identified by the Internet Watch Foundation
People are asking
**What does the 400,000 figure count?
** Reports to NCMEC’s CyberTipline involving generative AI during 2025. Its annual data separately identifies more than 182,000 reports involving AI-generated child sexual abuse material.
**How does generated material harm real children?
** Offenders alter existing abuse imagery to create new content and repeatedly victimise identifiable children, according to NCMEC.
**What safeguards is the IWF seeking?
** Testing before release, protections built into AI systems and strong moderation to prevent abusive content. These are its published proposals.
Elsewhere in AI
The US administration plans to designate Nvidia, Amazon, Microsoft, OpenAI and Anthropic among more than two dozen AI “national champions” to support overseas expansion, the Wall Street Journal reports. Read
The US and its allies have signed an accord to protect AI supply chains from China, ahead of Donald Trump’s meeting with Xi Jinping, the Financial Times reports. Read
This year’s ten biggest stock-market flotations have fallen more than 20% on average as the AI frenzy fades, the Financial Times reports. Read
