Weekly AI News Roundup – November 23, 2025
Six stories shaping the future: models, medicine, hardware, politics, and the rise of AI agents.
Artificial intelligence didn’t slow down this week — if anything, the velocity ticked upward. From Google’s newest model to worldwide commerce forecasts, this week’s developments showed a familiar pattern: AI is expanding in every direction at once, and the stakes (scientific, political, economic) are rising.
Here’s your full breakdown of the six most important AI developments this week, what they mean, and why they matter.
1. Gemini 3 Arrives — and Google Claims It’s Their “Most Intelligent Model Ever”
This week, Google DeepMind introduced Gemini 3, describing it as “our most intelligent model that helps you bring any idea to life.”
External source: Google announcement.
Gemini 3 represents Google’s strongest attempt yet to reclaim center stage in the model-capability race. While the headline feature is increased reasoning power across text, images, and video, the underlying message is clear: multimodality is no longer optional — it’s the baseline for next-gen models.
Google claims Gemini 3 can unify several steps of creation and analysis into one thread. Instead of “upload → analyze → refine,” the model is designed to collapse those steps into a single flow. For developers and businesses, that’s important: AI is trending toward end-to-end agents rather than tool-based workflows.
What this means for users & businesses
- Expect a wave of new products built atop Gemini 3 once Dev tools unlock.
- This fuels the question your audience keeps asking: Are we approaching AI that can fully handle multi-step tasks with minimal human guidance?
- Strong material for your upcoming evergreen article on multimodal AI — insert internal link here.
2. AI vs Superbugs — GSK & Fleming Initiative Launch £45M AMR Research Program
External source: GSK announcement.
In one of the biggest AI-for-science announcements this fall, GSK and The Fleming Initiative unveiled six new programs using AI to fight antimicrobial resistance (AMR) — particularly deadly pathogens like MRSA and Gram-negative bacteria.
This isn’t the “AI cures everything” hype cycle. It’s a targeted push where AI models help simulate protein structures, analyze antibiotic pathways, and identify new compounds faster than traditional lab cycles.
Why this matters globally
AMR is projected to kill more people annually than cancer by 2050 if nothing changes. Using AI to accelerate compound discovery is one of the few paths that may shift that trajectory.
3. U.S. House Republicans Move to Block State-Level AI Regulations
External source: Axios report.
AI regulation heated up this week as House Republicans drafted a proposal that would override individual state AI lawsby inserting federal language into the National Defense Authorization Act.
If passed, this would centralize AI governance at the federal level and prevent states from rolling out their own rules — such as California’s deepfake restrictions or New York’s biometric requirements.
The underlying conflict
This is becoming a classic U.S. policy battle:
- States want strict, granular controls.
- Federal lawmakers want a single, unified playbook.
- Tech companies want clarity — but minimal constraints.
Why this matters
AI regulation is repeatedly hitting the same friction point as privacy, cybersecurity, and big-tech antitrust: fragmented rules slow down innovation, but federal pre-emption sparks political wars.
see: AI Regulation Is Coming: What It Means for Businesses
4. IBM & University of Dayton Announce Joint Semiconductor Research Facility
External source: IBM newsroom.
IBM and the University of Dayton revealed plans for a next-generation semiconductor and packaging research center slated to open in 2027, focusing on:
- AI-optimized chips
- Photonic interconnects
- Advanced packaging
- Ultra-dense compute modules
Why this is important
Behind every AI breakthrough — GPT-4, Claude, Llama, Gemini — there’s an invisible constraint: hardware limits are becoming the real bottleneck.
This partnership signals:
- Universities are taking a larger role in national AI infrastructure.
- The U.S. is trying to close its gap in chip production and research.
- Compute scarcity isn’t going away soon.
5. AI Agents Expected to Drive 22% of Global Cyber Week Sales
External source: Retail TouchPoints forecast.
One of the most eyebrow-raising stats this week:
AI agents are projected to influence 22% of all global online sales during Cyber Week — approximately $73 billion.
This includes agents that:
- compare prices
- handle customer support
- auto-fill shopping carts
- make purchase recommendations
- complete checkout flows
Why this matters
This is a dramatic sign that AI has quietly embedded itself in commerce infrastructure. And the story isn’t “AI agents buying things on their own” — it’s that retailers are using agentic systems everywhere in the customer journey.
For businesses, this signals a shift: if you’re not using AI to support customer workflows, you may fall behind competitors who are.
6. Deloitte’s 2026 TMT Forecast: The Gap Between AI Promise & Reality Narrows — But Persists
External source: Deloitte Insights.
Deloitte’s annual Tech/Media/Telecom predictions arrive with a familiar but important message:
AI hype is still outpacing AI adoption — but the distance between them is tightening.
While nearly every company claims to be using AI, most are stuck at pilot stage. Only a fraction reach production-scale deployment.
Why this matters now
It sharpens a truth that companies are quietly wrestling with:
AI is becoming indispensable, but integrating it into real workflows remains harder than the marketing implies.
🧠 The Big Picture This Week
Across all six stories, a pattern emerges:
- Models are getting stronger (Gemini 3)
- AI is moving deeper into hard science (AMR research)
- Regulators are circling (federal vs state battle)
- Infrastructure is becoming strategic (IBM/Dayton chip center)
- Commerce is shifting (AI agents handling billions in sales)
- Businesses are waking up to reality (Deloitte: scaling remains difficult)
The through-line?
AI isn’t just advancing — it’s entangling itself into every critical system: medicine, politics, industry, and economics.
And that sets the stage for a very different 2026: more regulated, more multimodal, more agent-driven, and more hardware-dependent.
