The Neural Link | Edition 12

Welcome to the latest edition of The Neural Link! 

In this edition, we track the biggest leaps and shake-ups from the past month.  From Gemini 3’s headline-grabbing debut and OpenAI’s race to catch up, to Samsung’s tiny model outperforming the giants and Australia’s bold AI rollout in government, the momentum isn’t slowing.

We’ve also rounded up the latest shifts in AI regulation, infrastructure, music licensing, and media, plus a few surprises you might not have seen coming.

Let’s dive in.

This month saw a flurry of new model releases, from flagship upgrades to frontier agents reshaping how AI works.

Australia rolls out AI for public sector

The Australian government has announced a national plan to train all federal public servants in generative AI. The decision follows internal trials with Microsoft Copilot, which showed productivity gains, even as some concerns were raised about data handling and security.

Why it matters:
This is one of the most ambitious public-sector AI rollouts globally. It signals a shift in how governments plan to use gen AI,  not just to test it, but to embed it into everyday work.

Read the full story

Study finds 95% of firms see no AI ROI

An MIT study found that 95% of companies using generative AI haven’t yet seen a return on investment. Despite broad interest, only a small number of organisations are deploying AI at scale. The report highlights a gap between experimentation and measurable impact.

Why it matters:
While AI pilots are everywhere, proven results are rare. This is a wake-up call for business leaders to focus on use cases that move the needle and to plan for scale, not just proof-of-concept.

Read the full story

Tencent predicts AI will create 30% of films by 2026

The chief of Tencent Video predicts that by next year, generative AI could produce up to 30% of long-form film and animation content. Chinese studios are already adopting AI tools for storyboarding, voice, animation, and editing at a rapid pace.

Why it matters:
AI-generated content is no longer a fringe experiment; it’s becoming a mainstream production tool. If Tencent’s forecast is accurate, the entertainment industry is about to face a serious creative and economic shake-up.

Read the full story

Warner Music signs deal to allow AI vocals

Warner Music has signed a landmark agreement with AI music startup Suno, settling a lawsuit and allowing fans to create songs using the voices of select artists. The deal introduces new licensing terms designed to monetise AI-generated music.

Why it matters:
Rather than fighting AI-generated music, Warner is finding ways to commercialise it. This could set a precedent for how labels approach vocal cloning, fan-generated tracks, and new revenue streams.

Read the full story

Samsung’s tiny model beats giants on logic

Samsung researchers have developed a 7-million-parameter “Tiny Recursive Model” that outperforms large language models more than 10,000× its size on logic and reasoning tasks. The breakthrough highlights the power of architectural innovation over brute-force scale.

Why it matters:
This flips the script on the bigger-is-better model race. Efficient, smaller models could make powerful AI more accessible, especially in low-power and edge computing environments.

Read the full story

Microsoft 365 Copilot is now available for SMBs

Microsoft has launched Copilot for Business- a new AI offering tailored to SMBs with up to 300 users. It delivers the same powerful AI features as the Enterprise Copilot SKU but at a lower price point ($21/user/month) and with simpler licensing. The tool integrates directly into familiar apps like Word, Excel, Outlook, and Teams, helping smaller teams automate admin, draft content, and analyse data without the complexity.

Why it matters:
This is the first time enterprise-grade AI productivity has been made accessible to small and mid-sized businesses. It’s a practical way to put AI to work across day-to-day operations- no custom tools, no retraining required. 

Read the full story

Other news in AI

• EU delays AI rules, draws backlash
  The European Commission proposes pushing strict AI regulations to 2027, sparking criticism over favouring Big Tech.

• ChatGPT adds Group Chat for teams
  OpenAI now lets users collaborate in real-time with ChatGPT via its new Group Chat feature.

• Anthropic announces $50B US data centre plan
  Claude’s maker plans massive infrastructure expansion to support US-based AI leadership.

• Australia pushes for AI content labelling
  New guidance urges watermarking of AI-generated content, ahead of a national deepfake law rollout.

• AI-generated hit pulled over vocal cloning
  A viral track by Haven was removed after mimicking a real artist’s voice, prompting tighter label-AI partnerships.

• Calls to ban OpenAI’s Sora 2 video model
  Advocates say hyper-realistic deepfakes from Sora 2 outpace safety efforts and threaten public trust.

• IBM: It’s your data, not your model
  An IBM survey of 1,700 execs finds that messy, siloed data is the top barrier to enterprise AI success.

• Meta may switch to Google AI chips
  Meta is in talks to buy billions of Google TPUs for its data centres, a potential shakeup in the AI chip wars with Nvidia.

That’s a wrap on 2025. We’ll be back in the new year with more insights, bigger stories, and a closer look at what’s next.

Want to see what AI can do for your business? Get in touch with our experts to explore solutions built for your goals, not just the hype.

And if you haven’t already, subscribe to The Neural Link for a monthly dose of insight into the trends that matter in AI and automation.