From Gemini 3 to Federal AI Policy — What Actually Matters This Month

In this edition: it’s less about flashy demos and more about integration, readiness, and power consolidation. We’re seeing better tools — and stronger government intervention.

This is Q&AI, our blog series aimed at keeping you in the know on updates in the rapidly evolving world of AI. Sometimes, these will be quick updates on new developments in the field. Sometimes, they’ll be tips on tactics, features, or functionality. If you haven’t met me yet, hi: I’m Jen Taylor, CI’s Director of AI Strategy & Implementation, and your (very human) AI BFF. 

Q: WHat changed in U.S. AI policy? 

A: The Trump administration issued an executive order in December asserting federal authority over AI regulation, overriding state-level efforts.

The upside:

• A unified national standard
• Less regulatory fragmentation for companies

The concern:

• Centralized control may prioritize speed and deregulation
• Consumer protections could take a back seat

This is a pivotal moment, and one that will shape how fast AI advances…and who bears the risk.

Q: Quick one! What’s the deal with using quotation marks when giving information to ChatGPT?

I find myself always typing: and here’s the approved blurb: “copy copy copy” and here’s the artist’s bio: “copy copy copy? Do I need to bother with the quotation marks? It’s definitely a habit, from using them in Google searches.

A: No need! I usually do something like, here’s the content: and then use a ; between sections. It will understand however you want to format it—nothing specific recommended.

Q: Does anything stand out in the latest model updates? 

A: See my breakdown below!

From Google: Google launched a new Gemini 3 model in mid-December, and the standout is the combination of strong reasoning with real speed. That balance is hard to pull off (models usually trade one for the other).

I’ve been increasingly impressed with Gemini over the past few months, and this update reinforces the sense that Google is no longer just “catching up.” I’m excited to test this model more deeply in real workflows, not just benchmarks.

Google also released Google Workspace Studio: a new no-code environment for connecting Gmail, Drive, and Chat into lightweight agents. The promise is big: letting non-engineers automate real workflows across Google’s ecosystem.

The reality (so far)? Conceptually exciting, but practically inconsistent.

It’s powerful infrastructure, but not yet reliable enough to sit on the critical path of daily work. Worth watching—but not betting your operations on.

From OpenAI: OpenAI’s updated image model claims 4× faster generation speeds and more precise editing, meaning you can change specific elements without regenerating the entire image.

This is clearly aimed at solving the “slot machine” problem of AI art—where every prompt tweak produces a totally different result. It’s also a direct response to Google’s Nano Banana image model.

This feels less like an art play and more like a move toward usable, repeatable creative workflows.

From Anthropic: Anthropic has an internal research tool called Interviewer, designed to plan and conduct structured interviews with humans at scale. It’s not public…but it’s a signal.

This points to the rise of agentic research: AI systems that don’t just analyze data, but actively gather qualitative insight. If this works, it could make deep, human-centered research faster and cheaper than traditional surveys—fundamentally changing how organizations learn.

MORE SOON,
Jen

Have a question for a future edition? Submit it here!