I’ve described 2025 as my year of AI. And what a year it’s been.
But in today’s long-listen, I’m looking back at the year of AI in general, the news, views, and big trends.
If 2024 was the year AI became a household topic, 2025 was the year it turned into a hard reality. Billions were invested, billion-user platforms pushed AI deeper into everyday tools, and regulators finally brought rules with real force. China shifted expectations overnight. Coding started changing at a speed. And the gap between hype and what AI can actually do began to close.
With almost 270 episodes of AI SIX this year, I’ve tracked every twist and turn of this extraordinary 12 months. Today, I’m going into my podcast archives, bringing you the AI year in review. The stories that shaped AI in 2025, and what they mean for all of us heading into 2026.
The DeepSeek Earthquake
Everything changed in January.
On 20 January 2025, a relatively unknown Chinese AI lab released a model called DeepSeek R1. Within a week, almost $600 billion was wiped off Nvidia’s market value.
Why did it matter?
For years, Silicon Valley believed that frontier AI success required massive scale: more compute, more GPUs, more money. DeepSeek shattered that assumption.
- Reported training cost: $6 million
- Comparable Western models: $100 million+
- Performance: on par with OpenAI’s o1 on reasoning tasks
- And crucially: open source
The reaction was immediate. Sam Altman called it a wake-up call. Tech stocks dipped. And the narrative shifted from “scale above all else” to “efficiency matters too.”
For organisations outside frontier labs, the message was clear: you don’t need Google-sized budgets to build useful AI. That theme comes up again and again in 2025.
Regulation Arrives
February marked another turning point.
On 2 February 2025, the first prohibitions under the EU AI Act came into force. For the first time, specific AI uses were outright banned across all 27 EU member states.
These included:
- Social scoring systems
- Facial recognition databases built from scraped images
- Emotion recognition in workplaces and schools
- Predictive policing based purely on profiling
The penalties?
Up to €35 million or 7% of global annual turnover, whichever is higher.
This wasn’t just European bureaucracy. It was the first comprehensive AI law anywhere in the world, with extraterritorial reach. If your AI system touches EU citizens, it applies to you.
By August, the rules extended to general-purpose AI models, introducing transparency and copyright obligations. 2025 was the year AI regulation stopped being theoretical and started being enforceable.
The Rise of Vibe Coding
One of my favourite moments of the year came from a single tweet.
In February, AI researcher Andrej Karpathy coined the term “vibe coding”: building software by describing what you want in plain English and letting AI write the code.
I explored this in Episode 254 when I vibe-coded my own app. Karpathy wasn’t exaggerating.
By March:
- 25% of Y Combinator startups reported codebases that were 95% AI-generated
- Non-developers were suddenly building real tools
- Prototyping became radically faster
Is vibe coding production-ready? Not always. Security, bugs, and technical debt remain real concerns. But for experimentation and innovation, it’s transformative.
Collins Dictionary named “vibe coding” Word of the Year 2025—and rightly so.
The Year of AI Agents
If there’s one technical theme that defined 2025, it’s this:
AI stopped answering questions and started taking action.
We moved from chatbots to assistants to agents.
- Chatbots respond
- Assistants help
- Agents act autonomously
Throughout the year, we saw:
- Anthropic’s Model Context Protocol (MCP) become a standard
- OpenAI’s Atlas browser, where AI navigates websites on your behalf
- Google’s Agent2Agent protocol enabling AI collaboration
- Claude Code demonstrating extended, autonomous problem-solving
By year’s end, 82% of executives said they plan to adopt AI agents within one to three years.
For marketing and communications teams, agents aren’t futuristic—they’re operational. The real question now is how to use them responsibly.
The Model Wars Heat Up
Competition between AI giants intensified throughout 2025.
OpenAI began the year strong with:
- o1 reasoning models
- ChatGPT Pro
- Sora video generation
But August brought a wobble. The much-hyped GPT-5 landed to a muted response, opening the door for Google.
And Google took it.
With Gemini 3 launching in November, Google caught up—and in some benchmarks, pulled ahead. By year’s end, the narrative had flipped completely.
OpenAI responded with GPT-5.2, introducing multiple variants and massive context windows. The takeaway for users? Competition is working. Models are faster, cheaper, and better.
AI Video Grows Up
AI video took a huge leap forward in 2025.
- Sora 2 introduced synchronised dialogue and native audio
- Google Veo 3 and Kling 2.6 raised the bar
- The quality gap between AI video and professional production shrank fast
Small businesses began producing ads independently. Global brands followed. But copyright tensions exploded—culminating in Disney’s $1 billion investment in OpenAI, licensing over 200 iconic characters for AI generation.
This wasn’t just a deal. It was a signal that intellectual property itself is being renegotiated in the AI era.
The Great AI Hype Correction
MIT Technology Review called 2025 “The Great AI Hype Correction.”
They weren’t wrong.
- 95% of businesses reported no value from early AI trials
- AI agents struggled with basic workplace tasks
- AI-related incidents rose sharply
But this wasn’t failure—it was maturity.
We moved from “AI can do anything” to “AI can do specific things well.”
That’s where real value lives.
AI at Work
AI adoption in the workplace accelerated rapidly:
- 88% of organisations now use AI in at least one function
- Costs dropped ~30% annually
- Enterprise use of ChatGPT crossed one million accounts
And yet, trust remains fragile.
Workers want help with repetitive tasks—but protect creative and relational work fiercely. The message from multiple studies was clear: AI is welcome as a collaborator, not a replacement.
The human-in-the-loop isn’t optional. It’s essential.
Looking Back, Looking Forward
2025 was the year AI stopped being optional.
For me personally, it meant:
- Training over 3,000 people
- Delivering workshops across Ireland, the UK, and beyond
- Speaking on 23 stages in 17 cities
- Recording 260+ podcast episodes
What I’ve learned is simple:
The organisations that succeed with AI aren’t the ones with the biggest budgets. They’re the ones with clarity, discipline, and governance.
2025 was the year of AI engagement.
2026 must be the year of AI impact.
The tools are ready.
The rules are forming.
Now the question is whether we’re ready to use AI wisely.
Thanks for listening to this special year-in-review episode of AI SIX.
And remember: Know it. Use it. Govern it.
About the Podcast Host
AI and Communications Specialist | Author | Podcaster | International Speaker
Joanne Sweeney is a recognised leader in digital marketing and digital transformation, specialising in the practical application of AI to enhance communication strategies. The founder of Digital Training Institute, Joanne, empowers leaders and communications professionals to leverage AI responsibly and effectively, ensuring transparency, trust, and tangible results.
She completed the Harvard Kennedy School Executive Programme in AI and Public Policy this summer before launching her fourth book AI Playbook, in September. Joanne has also published two editions of Public Sector Marketing Pro, which were preceded by her first title, Social Media Under Investigation Law Enforcement and the Social Web.
Her books have become go-to resources for governments, public sector and policing agencies, as well as marketing and communications professionals, all navigating the challenges of the changing Internet.
Joanne is the host of the AI SIX Podcast, a daily six-minute podcast which provides actionable insights, expert interviews, and practical advice to help professionals navigate the new world of work.
