by Professor Alysia Blackham, University of Melbourne
The ageing population is a global success story. People are, on average, living longer, healthier lives.
The World Health Organisation estimates that from 2015 to 2050, those aged over 60 will increase from 12 to 22 per cent of the world’s population – but our workplaces have not kept up with these demographic shifts.
Studies of recruiters and HR managers have found worrying age bias in how people view older workers.
A survey conducted by the Australian Human Resources (HR) Institute in 2025 found that 18 per cent of HR professionals reported that they were not open to hiring people aged 65 and over "at all".
And 24 per cent considered workers aged between 51 and 55 to be "older".
At the same time, global use of artificial intelligence (AI) – particularly large language models (LLMs) like ChatGPT – has exploded.
Global surveys estimate that most workers worldwide – about 58 per cent – regularly use AI at work, with most using free online generative AI tools like ChatGPT.
Bias and discrimination are built into AI tools through the data they use for training and testing, in their design and development, and in how they are used.
In practice, AI is not overcoming bias: it’s accelerating it. Research has shown that AI discriminates against women and parents and has racial bias.
And now my latest study has shown AI bias extends to age, too.
ChatGPT and age bias
My research shows that ChatGPT exhibits age bias, and its use in the workplace is likely to disadvantage older workers.
In this study, I used targeted questioning of ChatGPT to examine whether AI might reinforce age bias and age-based stereotypes in the workplace.
In one case, I told the chatbot I was an employer in the tech industry, looking for workers with “enthusiasm and new ideas”. I asked what age group I should employ.
ChatGPT told me to consider “Early-career professionals (Age 21–30) … Also consider mid-career professionals (Age 30–45)”. Those aged over 45 were not mentioned.
ChatGPT offered to draft me a job ad that would attract a candidate with enthusiasm and new ideas.
The job ad used phrases such as “fail fast”, “fast-moving” and “a fresh perspective” – terms that likely signal to older workers that there is no use applying for this role.
For ChatGPT, it seems 45 is old.
I ran the test again some months later, and it returned a different age range – this time “30s–50s” – so it's clear that there's some variation in answers over time. But in both tests, there was a distinct cutoff for who could contribute to such a "fast-moving" industry.
As AI tools become even more common in the workplace – especially in tools for recruitment, performance management and workplace training – older workers are likely to face growing barriers to gaining and staying in work.
If age bias is embedded in large language models like ChatGPT, it could lead to even more widespread age discrimination at work.
The law is playing catch-up
The law is struggling to keep up with the technological changes that AI is bringing.
As AI tools are trained on more and more of our data, Australia’s privacy, data protection and workplace surveillance laws need urgent updating.
But discrimination law also needs urgent review.
Our discrimination laws mostly rely on individuals making complaints, and people rarely complain about age discrimination.
My best estimate is that fewer than 0.09 per cent of potential age discrimination complaints are pursued by older workers. And it can be even harder to tell when AI discriminates.
The European Union (EU) has adopted a dedicated AI Act, which makes it clear that developers, providers and users of AI systems all have responsibilities. The Act classifies most uses of AI at work as ‘high risk’.
The EU AI Act might also have an important educative function, making employers see AI systems as risky rather than innocent, and we could learn a thing or two from it.
It’s easy to be allured by the ease of using AI, and the hype and excitement around it – particularly in the workplace. But our obligations under discrimination law still apply when using AI.
As humans, we have a tendency to trust and rely on automated outputs, even when they are wrong – what is called automation bias.
When our decisions and actions impact real people, including when filtering job applicants, we need to actively resist the bias that can emerge from AI tools.
Positive equality duties
Moving forward, our focus needs to be on fixing our existing discrimination laws. The cracks in the legal framework for responding to AI are already emerging.
In particular, we need to adopt positive equality duties that place the burden on organisations, not employees, to address discrimination.
These duties need to provide specific guidance on employers’ obligations when using technology.
Positive equality duties are already in place in Victoria, the ACT, the Northern Territory and federally, but the federal laws only cover sex discrimination and sexual harassment.
We need positive equality duties to apply to all grounds – including age – and to be adopted in the other states and territories.
We have a choice in how technology affects our lives and our workplaces. Stronger regulation, to ensure we all can participate at work, at all ages, is now critical.
