By Jennifer Dudley-Nicholson

Australians can expect to use more artificial intelligence tools in 2025, according to tech experts, but the software will also become smarter, more regulated, and less likely to make things up. 

The forecasts come after a record-breaking year for generative AI tools, which saw the software introduced to smartphones, televisions, vehicles and social networks, and increasingly welcomed in universities, workplaces and government agencies.

But with the value of AI expected to soar to $US826 billion by 2030, up from $US184 billion in 2024 according to Statista, big changes are on the way.

These are five AI trends to watch in the coming year, according to Australian academics and industry experts.

Agents of change

While generative AI tools first appeared as chatbots in late 2022, the technology will power “agents” in 2025.

In launching Gemini 2.0 in early December, Google chief executive Sundar Pichai said AI agents would be able to perform a series of tasks automatically and “bring us closer to our vision of a universal assistant”.

“We have been investing in developing more agentic models, meaning they can understand more about the world around you, think multiple steps ahead, and take action on your behalf with your supervision,” he said.

The tech giant is not alone in striving to create smarter AI, UNSW AI Institute chief scientist Toby Walsh said, but the challenge ahead would be substantial.

While AI chatbots answered questions or composed text, he said, AI agents would be tasked with making decisions and acting on them, creating legal and ethical challenges if something went wrong.

“It’s the real challenge of AI: systems that have some autonomy, some ability to act on their own in the real world,” he said.

“When you give it the ability to take control over parts of your computer or buy things and do things for you, the potential for things to go wrong is much greater.”

Strict rules 

The European Union’s AI Act will gradually come into force in 2025, with a ban on some high-risk AI uses in February, such as social scoring, and obligations on general AI systems enforced by August.

While some rules are already having an impact on internet users, University of the Sunshine Coast computer science lecturer Erica Mealy said its benefits were stacked in favour of European residents who were being offered the chance to opt in to AI features rather than having to opt out of them.

“If you’re not in the EU, too bad, so sad, we will make you opt out,” she said.

“That relies on you having more knowledge of the platforms and knowing what has been turned on.”

But Australia is likely to get its first legal restrictions on generative AI technology in 2025, with a parliamentary inquiry recommending the introduction of a law dedicated to regulating high-risk AI uses.

The Adopting Artificial Intelligence Inquiry also recommended any law include restrictions on general purpose AI tools, and that the government consult with creative workers on the use of their copyright material.

AI in everything

First Samsung, then Google, now Apple: all three major smartphone manufacturers have added generative AI tools to their devices.

The technology will enter even more technological spaces in 2025, Prof Walsh said, as manufacturers develop innovative ways to use the technology and embed it in all parts of an operating system, rather than keeping it in a single app.

“A bit of AI is going to be sprinkled into everything electrical,” he said.

“Every device and every workflow is going to have some AI sitting in the background, trying to help you out.”

Users can expect generative AI technology to enhance more photos and videos they capture, Prof Walsh said, to appear in more cars, on more online platforms, and embedded in more services.

Big business

While the technology is expected to enter a lot of pockets in 2025, AI is forecast to play an even bigger role in the business world.

The technology was named as the most revolutionary tech of 2025 by several Atlassian executives, and Team Anywhere head Annie Dean said it would continue to remove menial, routine tasks to help workers focus on larger projects.

“AI will erase busywork and give teams the exactly support they need to explore and execute big ideas, faster,” she said.

“When teams collaborate with AI, they can more easily bring their ideas to life.”

The federal government has also revealed plans to launch a National AI Capability Plan in the coming year to build on the $2 billion in venture capital invested in local AI applications in 2023.

The plan is expected to launch in late 2025, following industry and public consultations.

Foundational fixes

Investments in artificial intelligence are expected to keep growing in 2025, and some funds will be directed to fixing and improving foundation models.

While more work was needed to tackle hallucinations, in which AI tools generate incorrect or nonsensical information to fill in knowledge gaps, Prof Walsh said, future research would also focus on expanding their abilities.

“The foundation models are very good at retrieving information, summarising information, but they’re very bad at reasoning,” he said.

“There are plentiful examples where they will say one plus one is not the same as two.”

To build on AI models’ reasoning ability, researchers will need to train the technology in greater real-world context, such as gravity and rain, which programs are unable to experience.

“AI is advancing at an amazing rate and that’s because at the moment the world is spending a billion dollars a day on it,” he said.

“We are at the peak of the hype cycle and at some point the froth will blow off the top of the glass, but companies are still seeing a significant return.”