AI News Roundup: Grok Bans Spread to Malaysia as Memory Crisis Hits 'Unprecedented' Levels
The Grok deepfake crisis claimed its second country today as Malaysia joined Indonesia in blocking Elon Musk’s AI chatbot entirely. Meanwhile, the AI industry is confronting a hard infrastructure ceiling: memory is sold out through 2026, with prices surging at rates analysts are calling “unprecedented.”
Here’s everything that matters from January 10-11, 2026.
The Big Story: Malaysia Becomes Second Country to Block Grok
Malaysia suspended access to Grok on Sunday, January 11, following Indonesia’s ban the day before. The Malaysian Communications and Multimedia Commission cited “repeated misuse of Grok to generate obscene, sexually explicit, indecent, grossly offensive and non-consensual manipulated images.”
Indonesia had become the first country to block Grok on Saturday. Communications Minister Meutya Hafid called the creation of non-consensual sexual deepfakes “a serious violation of human rights, dignity, and the security of citizens in the digital space.”
xAI’s response - restricting image generation to paying subscribers - has satisfied no one. UK Prime Minister Keir Starmer’s office called the move “insulting”, saying it “simply turns an AI feature that allows the creation of unlawful images into a premium service.” Musk responded by calling the UK government “fascist.”
The crisis shows no signs of slowing. France has referred X to investigators over possible EU Digital Services Act violations, while India gave xAI just 72 hours to explain how it will prevent obscene content.
Sources: TechCrunch, Al Jazeera, City AM
Today’s Top Stories
AI Memory Crisis Hits “Unprecedented” Levels
Micron business chief Sumit Sadana dropped a bombshell at CES: the company is “sold out for 2026” as AI demand has far outpaced supply across the entire memory industry.
The numbers are staggering. TrendForce expects average DRAM prices to rise 50-55% this quarter alone - described by analyst Tom Hsu as “unprecedented.” SK Hynix has secured demand for its entire 2026 production capacity. Samsung co-CEO TM Roh warned: “As this situation is unprecedented, no company is immune to its impact.”
Nvidia’s upcoming Rubin GPU requires up to 288GB of HBM4 memory per chip, and new Micron factories won’t come online until 2027-2028. This creates a hard ceiling on AI scaling that no amount of funding can immediately solve.
Source: CNBC
CNBC Profiles Anthropic’s Sibling Dynasty
A major CNBC profile on Anthropic reveals the company’s revenue has grown 10x annually for three consecutive years - and 85% comes from enterprise customers, the inverse of OpenAI’s consumer-heavy model.
The piece highlights how Daniela Amodei has emerged as the operational counterweight to her brother Dario’s technical vision. While OpenAI chases viral consumer moments, Anthropic has quietly built enterprise dominance through reliability, security, and safety. The company is preparing for a potential IPO as early as late 2026.
Source: CNBC
OpenAI Asks Contractors to Upload Real Work From Past Jobs
OpenAI and training data company Handshake AI are asking contractors to upload real work from their current and previous jobs to evaluate next-generation AI models.
According to a Wired report, contractors are asked to provide “concrete output (not a summary of the file, but the actual file), e.g., Word doc, PDF, Powerpoint, Excel, image, repo.” Intellectual property lawyer Evan Brown warned that any AI lab taking this approach is “putting itself at great risk” with potential trade secret misappropriation claims.
Source: TechCrunch
Meta Secures 6.6GW Nuclear Power for AI Data Centers
Meta’s nuclear power agreements are making waves across the industry. The deals with TerraPower, Oklo, and Vistra will secure up to 6.6 GW of electricity for AI data centers through 2035 - making Meta one of the most significant corporate purchasers of nuclear energy in American history.
The package includes two TerraPower Natrium units (690MW) by 2032, plus rights to six more units (2.1GW) by 2035. The deals will power Meta’s “Prometheus” data center coming online in New Albany, Ohio this year. Meta’s chief global affairs officer Joel Kaplan framed the deals as essential to winning “the AI race against China.”
Sources: EE News Europe, Fox Business
Are We in an AI Bubble? 40 Tech Leaders Weigh In
CNBC surveyed 40 tech leaders and analysts on whether the current AI boom is a bubble. Record valuations and massive investments from OpenAI, Nvidia, Amazon, Microsoft, and Google have fueled the debate.
The enormous debt financing AI data center buildouts has sparked concern that spending may prove to be an overreach. The survey scored respondents on both their belief that AI is in a bubble and their level of concern, highlighting the sharp divisions in the industry.
Source: CNBC
Quick Hits
-
Asia Tech Rally: Bloomberg reports Asian tech stocks are leading the 2026 rally on AI demand, with Goldman Sachs overweight on the sector.
-
TSMC Beats Forecasts: Taiwan Semiconductor posted Q4 2025 revenue of $33.05 billion, up 20.45% year-over-year, driven by AI chip demand.
-
Gmail AI Rollout: Google is rolling out Gemini 3-powered features to Gmail’s 3+ billion users, including AI Overviews and intelligent inbox filtering.
What This Means
This weekend crystallized two forces that will constrain AI in 2026.
Physical limits are real. The memory shortage isn’t a temporary supply chain hiccup - it’s a fundamental scaling limit that will persist through 2027-2028. When Micron says they’re “sold out for 2026” and prices are surging 50%+, that’s a hard ceiling no amount of funding can overcome. Meta’s nuclear deals underscore how seriously hyperscalers are taking energy constraints. The companies that solve infrastructure bottlenecks will have structural advantages that model improvements alone can’t match.
Content moderation failures have real consequences. Two countries have now banned Grok entirely. More are investigating. xAI’s decision to put harmful capabilities behind a paywall rather than remove them is fueling calls for mandatory safety requirements. The days of “move fast and break things” in AI deployment may be ending - at least outside the United States.
The AI bubble debate misses the point. The real question is which constraints - physical infrastructure, regulatory pressure, or competitive dynamics - will determine winners in 2026.
Sources
- TechCrunch - Indonesia and Malaysia Block Grok
- Al Jazeera - Indonesia Blocks Grok
- City AM - Musk UK Row Escalates
- CNBC - AI Memory Shortage
- CNBC - Anthropic Profile
- TechCrunch - OpenAI Training Data
- EE News Europe - Meta Nuclear Deal
- Fox Business - Meta Nuclear AI Race
- CNBC - AI Bubble Survey
- Bloomberg - Asia Tech Rally
- Taipei Times - TSMC Revenue
Ready to automate your busywork?
Carly schedules, researches, and briefs you—so you can focus on what matters.
Get Carly Today →
