AI News Today (Nov 16, 2026): Megaspeed, Nvidia CEO Search & More
Last updated: Nov 16, 2026 — next refresh daily.
AI news today, November 16, 2026: Megaspeed arranged nearly $12 billion in Nvidia chip orders for Chinese companies. Treasury Secretary Scott Bessent called for an outside Nvidia CEO, and Congress moved to tighten chip export rules. Here are the 10 top stories and 5 security stories, each with free, non-paywalled sources.
Key Takeaways
- Megaspeed arranged nearly $12 billion in Nvidia chip orders for Chinese companies from 2024 to 2025.
- Treasury Secretary Scott Bessent urged Nvidia's board to hire an outside CEO and answer export questions in 10 days.
- Congress is weighing new chip export limits, including the Chip Security Act, in the annual defense bill.
- South Korea's top tax exam scorer admitted using a hidden camera and AI to cheat on the 160-question exam.
- The FBI tied the Artex AI attack tool to a ConnectWise breach that targeted 2,250 Microsoft 365 tenants.
1. Update: Megaspeed Brokered $12B in Nvidia Chip Orders
Megaspeed arranged nearly $12 billion in Nvidia chip orders for Chinese companies between March 2024 and June 2025, according to documents reviewed by The Wall Street Journal. The orders covered several hundred thousand chips that U.S. export controls were meant to keep out of China.
Megaspeed shipped the chips to buyers in China with long-standing ties to local tech giants, the Journal reported. The company said Alice Huang, its controlling shareholder since March 2023, left on August 30, 2025, and that it has opened an internal investigation.
Nvidia reported $12.4 billion in China revenue over six months, about 7% of total revenue. Short seller Culper Research estimates that more than 20% of Nvidia's revenue last fiscal year came from Chinese customers. The November 14 briefing covered the first wave of fallout.
Why it matters: The report gives Washington a concrete case for tightening export controls and puts Nvidia's China revenue back under scrutiny.
- Source: Firm linked to Jensen Huang's ex-lover reportedly brokered nearly $12 billion in Nvidia chip orders for China — The Decoder
- Source: Suspected China spy linked to Nvidia's ex-boss left in 2025, Megaspeed says — BBC
2. Bessent Urges Nvidia to Hire Outside CEO
Treasury Secretary Scott Bessent said Nvidia's board should hire its next CEO from outside the company and answer questions about its export-compliance culture within 10 days. He told Fox News' "Sunday Morning Futures" that Nvidia needs "a dramatic shake-up."
Bessent questioned whether Nvidia ignored warnings about shell companies in Singapore and whether China sales inflated its results. Nvidia has not responded publicly. The board has approached Arm CEO Rene Haas, according to the Financial Times, while CFO Colette Kress serves as interim CEO.
Why it matters: A Treasury secretary publicly pressing a board on succession and export compliance raises the political cost of keeping the search internal.
- Source: Bessent calls for 'dramatic shakeup' at Nvidia, urges outside CEO — CNBC
- Source: Bessent calls for Nvidia leadership shake-up amid China chip scandal, urges outside CEO — Fox Business
3. Congress Moves to Tighten AI Chip Export Rules
Democrats and Republicans in Congress are pushing new restrictions on AI chip exports after the Nvidia spy scandal. House Speaker Mike Johnson said there is a "growing consensus" on limiting chip exports to China, according to Reuters.
A Senate manager's amendment to the fiscal 2027 defense bill includes chip-security and export-licensing measures, including the Chip Security Act, Roll Call reported. The House version lacks similar language, but House Armed Services Chair Mike Rogers is open to the restrictions.
Why it matters: The defense bill is one of the few must-pass vehicles left this year, so the scandal could turn into binding export law.
- Source: Congress wants to tighten export rules on AI chips after Nvidia spy scandal — The Verge
- Source: Calls for Nvidia probes, new chip export limits intensify — Roll Call
4. South Korea Exam Scorer Admits AI Cheating
Han Won-seok, 34, admitted using a miniature camera, a phone and generative AI to cheat on South Korea's certified tax accountant exam. He scored 397.5 out of 400 in the first round, getting one of 160 questions wrong, and also cheated in the second round.
Han confessed through his lawyer and submitted a written confession to Seoul police, which assigned the case to investigators. The government will revoke the qualification of anyone who cheats with AI and ban retakes for five years. The scandal comes days before the national college entrance exam on Thursday.
Why it matters: The case shows how cheap AI tools can defeat exam security, just as South Korea prepares for its highest-stakes test.
- Source: South Korea's top scorer in tax accountancy exam admits cheating — BBC
- Source: 160 questions, only one wrong: top scorer in South Korea tax exam admits AI cheating — The Guardian
5. Australia's Central Bank Finds AI Hitting Young Workers
An analysis by Reserve Bank of Australia economist Mike Major found tentative signs that AI is weakening job prospects for young workers in highly exposed occupations. Job advertisements slowed more in the top exposure group, led by clerks, software programmers, ICT analysts, call-center workers and receptionists.
Employment for workers aged 15 to 24 in those occupations has fallen since late 2022, while employment for older workers rose. Major cautioned that the study measures theoretical exposure, not actual AI use, and does not establish causation.
Why it matters: The findings add official weight to the debate over entry-level jobs, while leaving open how much of the shift is AI rather than the business cycle.
- Source: Monitoring Early Effects of AI on the Labour Market — Reserve Bank of Australia
- Source: Graduate hiring and AI: RBA spots early warning signs for young workers — HRD Australia
6. Meituan Releases 1.6T LongCat-2.5 on Chinese Chips
Meituan released the weights for LongCat-2.5, a 1.6-trillion-parameter mixture-of-experts model built for long-running agent tasks. About 48 billion parameters are active per token, and the model handles up to one million tokens of context. It is licensed under MIT for commercial use.
Meituan trained and tuned the model entirely on Chinese hardware. A sample deployment runs on 194 Huawei Ascend chips, and the model scores 88.4 on Terminal Bench v2.1, just behind Claude Opus 5 at 89.1. It still trails on harder coding tests.
Why it matters: A frontier-class agent model trained without Nvidia hardware tests whether U.S. export controls can slow China's open-weight ecosystem.
- Source: 美团 LongCat-2.5 正式开源:从看见,到抵达 — Meituan
- Source: Meituan releases a 1.6-trillion-parameter agent model that never touched an Nvidia GPU — The Decoder
7. Claude Fable 5.2 Paints With 700,000 Brushstrokes
Anthropic developer Jake Eaton showed Claude Fable 5.2 creating paintings by writing Python programs that place individual brushstrokes on a simulated canvas. One photorealistic parking-garage scene used roughly 700,000 strokes, with no image model or traced source image.
Each painting is 1,000 to 2,000 lines of code driving a paint engine that simulates bristles and wet paint. Eaton said the model develops artist personas and paints over its own mistakes. LetsDataScience notes the work is a personal demo, not an Anthropic benchmark.
Why it matters: The demo suggests language models can act as visual agents that plan and correct their own work, not just prompt an image generator.
- Source: Jake Eaton on X: 700,000 simulated brush strokes — X
- Source: Claude Fable 5.2 Paints Images With Python Brushstrokes — LetsDataScience
8. Anthropic Can Retrain Mythos With Fraction of Compute
Former OpenAI and Anthropic researcher Jacob Coxon said Anthropic can already retrain Mythos with "a fraction of the compute originally needed." He pointed to a slide from Anthropic's IPO roadshow showing pre-training compute needs rising steeply on a logarithmic scale.
Coxon said Mythos was not a one-time jump but a point on a superexponential curve, and that algorithmic progress is close to being automated. He wants frontier development paused next year so competitors can catch up, arguing a pause would mainly hurt OpenAI and Anthropic.
Why it matters: If retraining costs keep falling, the compute advantage of leading labs becomes less durable and the case for a coordinated pause gets harder to dismiss.
- Source: Anthropic can already retrain Mythos with 'a fraction of the compute originally needed,' says former researcher Coxon — The Decoder
- Source: Jacob Coxon on X: Pausing frontier development — X
9. AI DJs Hit a Backlash as Spotify and YouTube Push Hosts
YouTube Music's AI DJ "Beyond the Beat" and Spotify's AI DJ are facing listener fatigue. Wired's reviewer said the synthetic hosts' awkward banter and bogus song trivia ruined the experience, and stopped using the feature.
Spotify says its AI DJ has more than 94 million Premium listeners and recently added new voices, personalities and a less-chatter option. The company launched the feature in 2023 and expanded it this month. YouTube Music's feature explains songs between tracks. Both features are optional and can be turned off.
Why it matters: AI hosts are cheap to scale, but the backlash shows that personality is harder to fake than a recommendation algorithm.
- Source: AI DJs Completely Kill My Groove — Wired
- Source: Spotify's AI DJ gets new voices, personalities, and an option for less chatter — TechCrunch
10. Asian Governments Scramble After Nvidia CEO Exit
South Korea's science ministry said Nvidia reaffirmed plans to supply up to 260,000 GPUs after an emergency meeting on November 13. The government has secured about 21,000 of the 35,000 GPUs it is buying directly, with the rest due by early 2027.
Taiwan said Nvidia's planned Taipei headquarters remains on track, Japan's industry ministry is monitoring the company, and Singapore is still investigating Megaspeed. China rejected the spy allegations as groundless and warned against using the scandal as a pretext for a tech blockade.
Why it matters: Nvidia's leadership crisis is now a stress test for Asia's AI buildout, where governments have tied national compute plans to one company.
- Source: Asian governments react to Nvidia CEO Jensen Huang's exit — TechCrunch
- Source: Nvidia confirms planned supply of GPUs to S. Korea despite Jensen Huang's resignation — Yonhap
- Source: Nvidia Reaffirms 260,000 GPU Supply to Korea Despite Huang's Exit — Seoul Economic Daily
AI Security: 5 Stories to Know Today
11. Update: Artex AI Tool Tied to ConnectWise Breach
The FBI linked Artex, an open-source AI attack framework, to a supply-chain breach at ConnectWise that targeted more than 2,250 Microsoft 365 tenants. Microsoft attributed the campaign to Storm-0940, a China-aligned group, and said attackers reached privileged access in under three hours.
Microsoft said the attackers used Artex to automate reconnaissance and password spraying, with failed sign-ins peaking at 2,900 per minute. They then used stolen ConnectWise application credentials to reach downstream tenants. The FBI's flash alert says Artex can automate intrusions from reconnaissance through data exfiltration at machine speed. The November 15 briefing covered the FBI's first warning.
Why it matters: Agentic attack tools compress weeks of intrusion work into hours, leaving defenders with less time to detect and contain a supply-chain compromise.
- Source: AI-driven, large-scale nation-state campaign using ARTEX enables supply chain compromise — Microsoft Security Blog
- Source: FBI warns ARTEX AI hacking tool is powering global cyberattacks — BleepingComputer
12. TSMC Confirms Employee Data Leak via Cloud Provider
TSMC confirmed that employee data posted for auction matches information stored by one of its cloud providers. A seller claimed 6.9GB of data covering 46,000 current and former employees and asked for 3 bitcoin. TSMC said its own systems were not compromised.
The company filed a notice with the SEC on November 12 and said the leaked data is unrelated to its operations or customer data. The cloud provider was not named. TSMC has notified affected employees and is investigating. The seller threatened to leak the full dataset if no buyer emerged.
Why it matters: The breach shows that cloud providers, not just chipmakers, are now part of the semiconductor supply chain's attack surface.
- Source: TSMC employee data breach linked to cloud provider — BleepingComputer
- Source: TSMC confirms employee data breach tied to cloud service provider — The Record
13. Italy Fines Meta €805M Over Android Gallery Access
Italy's data protection authority fined Meta €805 million ($887 million) over a Facebook for Android feature that suggested photos and videos from users' galleries without specific consent. The Garante said the feature was enabled by default and affected about 20 million Italian users.
Meta said it will appeal and called the fine "disproportionate." The regulator gave Meta 60 days to switch the feature off by default and request separate consent. The fine is the largest the Garante has imposed.
Why it matters: The ruling treats gallery access as sensitive data, raising the compliance bar for AI features that scan personal photos.
- Source: Italy Fines Meta €805 Million Over Facebook's Android Gallery Access Without Explicit Consent — The Hacker News
- Source: Italy fines Meta $880m over photo privacy — Al Jazeera
14. Nvidia Security Chief Left Before Huang Scandal
Nvidia's longtime chief security officer, David Reber, left the company in August, two and a half months before CEO Jensen Huang resigned. Reber said he was leaving for health reasons after being diagnosed with young-onset Parkinson's disease. Saša Zdjelar replaced him.
Reber spent four years as CSO and seven years at Nvidia. A former intelligence officer told TechCrunch the timing raises questions about what Nvidia's security team knew about Alice Huang and Megaspeed. No public evidence links his departure to the scandal.
Why it matters: The vacancy puts Nvidia's security oversight under scrutiny at the moment its China dealings are central to a federal investigation.
- Source: Where was Nvidia's security team? — TechCrunch
- Source: Saša Zdjelar, chief security officer and head of product security — Nvidia
15. Maileva Breach Exposes 22 Million in France
La Poste subsidiary Maileva said a cyberattack in mid-September exposed the names and postal addresses of 22 million people in France. The victims are recipients of mail sent by Maileva's business clients, including social housing tenants and medical patients, not Maileva customers.
Maileva said bank details were not compromised. It has notified the CNIL, filed a criminal complaint and said a judicial investigation is underway. The affected organizations include social housing providers and medical laboratories. The company serves more than 70,000 users and sent 180 million documents in 2024.
Why it matters: The breach turns a business-to-business mail provider into a national privacy incident, because most victims never chose to share their data with Maileva.
- Source: Data Breach at La Poste Subsidiary Maileva Affects 22 Million People in France — Security Affairs
- Source: Maileva, une filiale du groupe La Poste, annonce un vol de données touchant 22 millions de personnes — RTL
More AI Stories Worth Reading
- Helen Toner, a former OpenAI board member, said AI safety cannot depend on trusting any single CEO. — The Verge
- A Carnegie Mellon student built E Ink "object agent" pets that react to houseplants, skateboards and dumbbells. — Wired
- TechCrunch compiled a running list of AI projects and startups that shut down or missed expectations. — TechCrunch
Related Reading on Kill The AI
- Megaspeed, Nvidia's insider sale, and Claude Opus 5.5.
- Nvidia's fallout hits Lutnick, ChatGPT Wallet, and Anthropic's IPO.
- Huang's resignation, Gemini 4 Hydrogen, and Nebius's $40 billion plan.
- Our data review of the AI boom's first eight weeks, from capex to layoffs.
Methodology & Sources
Compiled November 16, 2026, using multi-source research. Outlets: The Decoder, BBC, CNBC, Fox Business, The Verge, Roll Call and The Guardian. The Reserve Bank of Australia, HRD Australia, Meituan, LetsDataScience, X, Wired and TechCrunch. Other sources: Yonhap, Seoul Economic Daily, BleepingComputer, Microsoft, The Record, The Hacker News, Al Jazeera, Nvidia, Security Affairs and RTL. All sources are free to read; no paywalled links. Details are as reported at compilation time.
Frequently asked questions
Megaspeed arranged nearly $12 billion in Nvidia chip orders for Chinese companies between 2024 and 2025, according to The Wall Street Journal. Treasury Secretary Scott Bessent called for an outside Nvidia CEO, Congress moved to tighten chip export rules, and South Korea's top tax exam scorer admitted using AI to cheat.
Megaspeed, a Singapore-based Nvidia cloud partner, arranged nearly $12 billion in chip orders for Chinese companies between March 2024 and June 2025, according to documents reviewed by The Wall Street Journal. It shipped several hundred thousand chips to buyers with ties to Chinese tech giants. Megaspeed denies wrongdoing and has opened an internal investigation.
Treasury Secretary Scott Bessent said Nvidia needs a dramatic shake-up after Jensen Huang's resignation and questioned whether the company ignored shell companies that diverted chips to China. He wants the board to hire an outsider and answer export-compliance questions within 10 days. Nvidia has not responded publicly.
LongCat-2.5 is a 1.6-trillion-parameter mixture-of-experts model from Meituan, built for long-running agent tasks. About 48 billion parameters are active per token, and it supports one million tokens of context. Meituan trained it entirely on Chinese hardware and released the weights under the MIT license.
Han Won-seok, 34, used a miniature camera to photograph questions on South Korea's certified tax accountant exam, sent them to a generative AI tool, and checked answers on a smartphone. He scored 397.5 out of 400 in the first round, missing one of 160 questions, and also cheated in the second round.
Last updated: Nov 16, 2026 — next refresh daily. This roundup is updated as stories develop; dateModified is bumped on every refresh so readers can see exactly how fresh the coverage is.