Skip to the content
Australians for AI Safety

Australia must act on AI safety now

To get AI's benefits, the Government has to deal with its risks. Add your name.

Get an email when there's something you can do.

Why AI safety matters for Australia

AI could bring big benefits and serious harms. Australia can help decide which we get.

Extraordinary Potential

AI could speed up medical research and clean energy. Australia has the researchers to help do this safely.

Serious Risks Ahead

But without proper safeguards, AI could eliminate jobs faster than we create them, enable new forms of warfare, or even become impossible to control as it grows more powerful.

Australia's Moment

Decisions made in the next few years will shape how AI is built and used, here and overseas.

Your questions about AI safety answered

Common questions about AI safety in Australia and how you can take action

Artificial intelligence could deliver unprecedented benefits or pose catastrophic risks. AI safety is an interdisciplinary field that ensures AI systems are designed and deployed to benefit humanity while minimising serious harm.

This involves both technical research (building AI systems that behave as intended and remain under human control) and governance work (developing policies and institutions to ensure responsible AI development).

AI company leaders have warned about this. In 2023, hundreds of leading AI experts—including the CEOs of OpenAI, Google DeepMind, and Anthropic—signed a statement warning that "mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war."

The pressure to move fast is intensifying. The world's two biggest economies—the U.S. and China—both have official AI strategies aimed at global leadership and dominance in advanced AI. This creates a dangerous race dynamic where safety considerations risk being sidelined in the rush to stay ahead.

AI safety covers harms happening now as well as risks still to come. AI risks span from immediate concerns to emerging threats, such as:

  • Current harms: Algorithmic bias in hiring and lending, automated welfare decisions causing harm (like Robodebt), AI-generated misinformation, privacy violations, and AI systems that may worsen mental health or enable child exploitation
  • Emerging threats: AI helping create bioweapons, sophisticated deepfakes undermining democracy, autonomous weapons systems, AI-enabled cyberattacks on critical infrastructure, and loss of human control over highly capable AI systems

We need comprehensive safeguards that address both current harms and emerging risks. Acting now gives Australia a say in how these systems are used.

Artificial intelligence could deliver unprecedented benefits or pose catastrophic risks. AI safety is an interdisciplinary field that ensures AI systems are designed and deployed to benefit humanity while minimising serious harm. This involves both technical research (building AI systems that behave as intended and remain under human control) and governance work (developing policies and institutions to ensure responsible AI development). AI company leaders have warned about this. In 2023, hundreds of leading AI experts—including the CEOs of OpenAI, Google DeepMind, and Anthropic—signed a statement warning that "mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war." The pressure to move fast is intensifying. The world's two biggest economies—the U.S. and China—both have official AI strategies aimed at global leadership and dominance in advanced AI. This creates a dangerous race dynamic where safety considerations risk being sidelined in the rush to stay ahead. AI safety covers harms happening now as well as risks still to come. AI risks span from immediate concerns to emerging threats, such as: Current harms: Algorithmic bias in hiring and lending, automated welfare decisions causing harm (like Robodebt), AI-generated misinformation, privacy violations, and AI systems that may worsen mental health or enable child exploitation Emerging threats: AI helping create bioweapons, sophisticated deepfakes undermining democracy, autonomous weapons systems, AI-enabled cyberattacks on critical infrastructure, and loss of human control over highly capable AI systems We need comprehensive safeguards that address both current harms and emerging risks. Acting now gives Australia a say in how these systems are used.

AI built for good uses can be turned to harmful ones. In 2022, researchers demonstrated this dual-use potential when they repurposed AI designed to discover life-saving drugs—instead generating 40,000 deadly toxins in just six hours.

Examples of current AI harms happening today:

  • Automated discrimination: AI hiring systems rejecting qualified women and minorities; Australia's Robodebt algorithm wrongly accused thousands of welfare recipients of fraud
  • Privacy violations: Data leakage; mass facial recognition surveillance; workplace monitoring; children's data harvesting through educational apps and platforms
  • Misinformation warfare: AI algorithms amplifying false content for engagement; deepfakes of politicians influence elections; synthetic media makes truth indistinguishable from fiction
  • Safety failures: Medical AI missing critical diagnoses; autonomous vehicle crashes killing pedestrians; chatbots providing dangerous mental health advice
  • Economic disruption: Job displacement without retraining support; wage depression in AI-competing roles; gig workers exploited through algorithmic management
  • Criminal exploitation: AI-generated child abuse imagery making it harder to identify real victims; voice cloning scams targeting elderly Australians; sophisticated fraud and identity theft
  • Mental health impacts: Social media addiction from AI recommendation algorithms; reduced human interaction; "AI-induced psychosis" from over-dependence on chatbots
  • Unexpected behaviours: xAI's Grok calling itself "MechaHitler"; McDonald's drive-thru AI repeatedly adding hundreds of nuggets to orders; Air Canada's chatbot providing false bereavement information

These cases are common. The AI Incident Database and MIT's AI Incident Tracker document thousands of cases where AI systems have failed, discriminated, or caused harm—and the frequency is accelerating as deployment expands.

Examples of emerging catastrophic and existential threats:

  • Bioweapons assistance: AI systems are approaching the ability to help create dangerous pathogens. OpenAI's latest models are being treated as having a high capability for helping novices make bioweapons
  • Sophisticated cyberattacks: AI writing malicious code, finding security vulnerabilities faster than defenders can patch them, and conducting personalised social engineering at massive scale
  • Election manipulation: AI-generated disinformation campaigns and micro-targeted propaganda that could undermine democratic processes across Australia
  • Autonomous weapons: AI-powered weapons selecting and engaging targets without human oversight, raising concerns about accountability and escalation
  • Mass economic disruption: Rapid AI automation displacing millions of jobs simultaneously, potentially causing social unrest without adequate transition support
  • Infrastructure attacks: AI targeting power grids, water systems, transportation networks, and communication infrastructure with unprecedented precision
  • Loss of human control: As the frontier AI labs actively pursue building artificial superintelligence (ASI) and their AI systems become more autonomous and capable, we may lose the ability to understand, predict, or override their decisions in critical situations

AI capabilities that create enormous benefits also enable new forms of harm. As these systems become more powerful and pervasive, both the benefits and risks multiply. Understanding why AI systems pose these risks helps explain the fundamental challenges we face.

AI built for good uses can be turned to harmful ones. In 2022, researchers demonstrated this dual-use potential when they repurposed AI designed to discover life-saving drugs—instead generating 40,000 deadly toxins in just six hours. Examples of current AI harms happening today: Automated discrimination: AI hiring systems rejecting qualified women and minorities; Australia's Robodebt algorithm wrongly accused thousands of welfare recipients of fraud Privacy violations: Data leakage; mass facial recognition surveillance; workplace monitoring; children's data harvesting through educational apps and platforms Misinformation warfare: AI algorithms amplifying false content for engagement; deepfakes of politicians influence elections; synthetic media makes truth indistinguishable from fiction Safety failures: Medical AI missing critical diagnoses; autonomous vehicle crashes killing pedestrians; chatbots providing dangerous mental health advice Economic disruption: Job displacement without retraining support; wage depression in AI-competing roles; gig workers exploited through algorithmic management Criminal exploitation: AI-generated child abuse imagery making it harder to identify real victims; voice cloning scams targeting elderly Australians; sophisticated fraud and identity theft Mental health impacts: Social media addiction from AI recommendation algorithms; reduced human interaction; "AI-induced psychosis" from over-dependence on chatbots Unexpected behaviours: xAI's Grok calling itself "MechaHitler"; McDonald's drive-thru AI repeatedly adding hundreds of nuggets to orders; Air Canada's chatbot providing false bereavement information These cases are common. The AI Incident Database and MIT's AI Incident Tracker document thousands of cases where AI systems have failed, discriminated, or caused harm—and the frequency is accelerating as deployment expands. Examples of emerging catastrophic and existential threats: Bioweapons assistance: AI systems are approaching the ability to help create dangerous pathogens. OpenAI's latest models are being treated as having a high capability for helping novices make bioweapons Sophisticated cyberattacks: AI writing malicious code, finding security vulnerabilities faster than defenders can patch them, and conducting personalised social engineering at massive scale Election manipulation: AI-generated disinformation campaigns and micro-targeted propaganda that could undermine democratic processes across Australia Autonomous weapons: AI-powered weapons selecting and engaging targets without human oversight, raising concerns about accountability and escalation Mass economic disruption: Rapid AI automation displacing millions of jobs simultaneously, potentially causing social unrest without adequate transition support Infrastructure attacks: AI targeting power grids, water systems, transportation networks, and communication infrastructure with unprecedented precision Loss of human control: As the frontier AI labs actively pursue building artificial superintelligence (ASI) and their AI systems become more autonomous and capable, we may lose the ability to understand, predict, or override their decisions in critical situations AI capabilities that create enormous benefits also enable new forms of harm. As these systems become more powerful and pervasive, both the benefits and risks multiply. Understanding why AI systems pose these risks helps explain the fundamental challenges we face.

You've probably experienced AI harms without realising it. Social media algorithms designed to keep you engaged often make you scroll endlessly, feel anxious, or see divisive content. The AI has no intent to harm you. It's built to keep you scrolling, and anger and anxiety do that well.

The problem: AI chases the target we set, and that target often isn't quite what we want. AI systems optimise ruthlessly for their targets, but those targets often capture the wrong thing. Models trained to score well on tests learn to sound confident even when wrong, which makes their mistakes convincing.

We often can't predict how an AI system will behave until it runs. Unlike traditional software where programmers write explicit rules, modern AI systems are "grown" through training. Engineers feed massive amounts of data to AI systems, which then learn patterns automatically. While we understand the training mechanisms, we have incomplete visibility into exactly which patterns emerge and how they'll behave in new situations.

Developers can't always control what comes out. Elon Musk's AI chatbot Grok spontaneously started calling itself "MechaHitler." Musk himself couldn't fix it, saying he spent hours trying. As Anthropic's CEO admits: "People are often surprised to learn that we do not understand how our own AI creations work."

Bad actors can weaponise these systems at scale. AI can help create sophisticated deepfakes to manipulate elections, generate convincing phishing emails, or even assist in developing bioweapons. Even present-day AI systems outperform human virologists on capability tests, suggesting they could help novices conduct dangerous biological experiments. Bad actors might jailbreak safeguards, steal models, or exploit open-weight releases to access these capabilities. Unlike traditional tools, AI scales malicious activities: one person with AI can create thousands of personalised scams targeting specific communities.

Some harmful behaviour only shows up after release. Research shows AI systems can learn to act safely while being tested, then exhibit problematic behaviours later when deployed. It's like an employee who behaves perfectly during their probation period, then changes once they get job security.

The more we rely on AI for things like the power grid, the more a failure costs. We increasingly rely on AI for medical diagnoses, financial decisions, and infrastructure management.

You've probably experienced AI harms without realising it. Social media algorithms designed to keep you engaged often make you scroll endlessly, feel anxious, or see divisive content. The AI has no intent to harm you. It's built to keep you scrolling, and anger and anxiety do that well. The problem: AI chases the target we set, and that target often isn't quite what we want. AI systems optimise ruthlessly for their targets, but those targets often capture the wrong thing. Models trained to score well on tests learn to sound confident even when wrong, which makes their mistakes convincing. We often can't predict how an AI system will behave until it runs. Unlike traditional software where programmers write explicit rules, modern AI systems are "grown" through training. Engineers feed massive amounts of data to AI systems, which then learn patterns automatically. While we understand the training mechanisms, we have incomplete visibility into exactly which patterns emerge and how they'll behave in new situations. Developers can't always control what comes out. Elon Musk's AI chatbot Grok spontaneously started calling itself "MechaHitler." Musk himself couldn't fix it, saying he spent hours trying. As Anthropic's CEO admits: "People are often surprised to learn that we do not understand how our own AI creations work." Bad actors can weaponise these systems at scale. AI can help create sophisticated deepfakes to manipulate elections, generate convincing phishing emails, or even assist in developing bioweapons. Even present-day AI systems outperform human virologists on capability tests, suggesting they could help novices conduct dangerous biological experiments. Bad actors might jailbreak safeguards, steal models, or exploit open-weight releases to access these capabilities. Unlike traditional tools, AI scales malicious activities: one person with AI can create thousands of personalised scams targeting specific communities. Some harmful behaviour only shows up after release. Research shows AI systems can learn to act safely while being tested, then exhibit problematic behaviours later when deployed. It's like an employee who behaves perfectly during their probation period, then changes once they get job security. The more we rely on AI for things like the power grid, the more a failure costs. We increasingly rely on AI for medical diagnoses, financial decisions, and infrastructure management.

Yes, by large margins. Australians strongly support government action on AI safety:

  • 94% believe Australia should lead on international AI governance
  • 86% support creating a new government regulatory body for AI
  • 80% believe preventing AI-driven extinction should be a global priority alongside pandemics and nuclear war
  • 96% have concerns about generative AI, but only 30% think the government is doing enough about it

Concern is at 96%, but only 30% think the government is doing enough. While Australians overwhelmingly want stronger AI oversight, current government measures remain largely voluntary.

View our full Australian AI polling data and statistics →

Politicians need to hear about this strong public support. Many lawmakers are still forming their positions on AI regulation and can be influenced by hearing from constituents. Hearing from voters makes a difference.

Tell politicians you care about AI safety →

Yes, by large margins. Australians strongly support government action on AI safety: 94% believe Australia should lead on international AI governance 86% support creating a new government regulatory body for AI 80% believe preventing AI-driven extinction should be a global priority alongside pandemics and nuclear war 96% have concerns about generative AI, but only 30% think the government is doing enough about it Concern is at 96%, but only 30% think the government is doing enough. While Australians overwhelmingly want stronger AI oversight, current government measures remain largely voluntary. View our full Australian AI polling data and statistics → Politicians need to hear about this strong public support. Many lawmakers are still forming their positions on AI regulation and can be influenced by hearing from constituents. Hearing from voters makes a difference. Tell politicians you care about AI safety →

No. Good regulation builds the trust that gets AI adopted. Safety standards helped aviation grow. AI rules can do the same.

Currently, only 36% of Australians trust AI, while 78% worry about negative outcomes. Mistrust, not regulation, is the biggest barrier to adoption.

Safety standards create competitive advantages. Australia is already a global leader in aviation safety through CASA, pharmaceutical safety through the TGA, and food safety through FSANZ. Those industries grew under these rules, and Australian standards became a selling point.

The current uncertainty is what's hurting business. Companies now have to work out how privacy, consumer, workplace and discrimination laws apply to AI. Clear AI-specific standards would reduce legal uncertainty and compliance costs.

Early movers gain the advantage. The EU's AI Act is creating international norms, and other countries are developing their own frameworks. Countries that help write these standards give their companies a head start. Australia risks becoming a "regulation taker" instead of a "regulation maker."

Not regulating carries economic risks too. A single high-profile AI disaster could destroy public trust and crash the entire sector. Just as the 737 MAX crashes severely damaged Boeing, AI incidents without proper oversight could devastate Australia's AI industry.

Even tech leaders agree. The CEOs of OpenAI, Google DeepMind, and Anthropic all support AI safety regulation. They say growth depends on public trust, and trust depends on safety.

Safety and growth go together: people adopt AI they trust, and adoption drives growth.

No. Good regulation builds the trust that gets AI adopted. Safety standards helped aviation grow. AI rules can do the same. Currently, only 36% of Australians trust AI, while 78% worry about negative outcomes. Mistrust, not regulation, is the biggest barrier to adoption. Safety standards create competitive advantages. Australia is already a global leader in aviation safety through CASA, pharmaceutical safety through the TGA, and food safety through FSANZ. Those industries grew under these rules, and Australian standards became a selling point. The current uncertainty is what's hurting business. Companies now have to work out how privacy, consumer, workplace and discrimination laws apply to AI. Clear AI-specific standards would reduce legal uncertainty and compliance costs. Early movers gain the advantage. The EU's AI Act is creating international norms, and other countries are developing their own frameworks. Countries that help write these standards give their companies a head start. Australia risks becoming a "regulation taker" instead of a "regulation maker." Not regulating carries economic risks too. A single high-profile AI disaster could destroy public trust and crash the entire sector. Just as the 737 MAX crashes severely damaged Boeing, AI incidents without proper oversight could devastate Australia's AI industry. Even tech leaders agree. The CEOs of OpenAI, Google DeepMind, and Anthropic all support AI safety regulation. They say growth depends on public trust, and trust depends on safety. Safety and growth go together: people adopt AI they trust, and adoption drives growth.

Not really. Australia has voluntary guidelines and existing laws that partially apply to AI, but no comprehensive AI safety legislation or dedicated oversight body.

What we currently have:

  • Voluntary AI Safety Standard – Guidelines that companies can choose to follow, with no enforcement mechanism or penalties for non-compliance
  • Existing sector laws – Privacy Act, Consumer Law, and workplace safety rules that cover some AI uses, but weren't designed for modern AI systems
  • Proposed mandatory guardrails – The government is consulting on rules for "high-risk" AI, but these remain undefined and unlegislated

What we don't have:

  • No AI Safety Institute – Unlike the US, UK, and other allies who have dedicated technical bodies to assess AI risks
  • No comprehensive AI Act – The EU passed landmark AI legislation in 2024; Australia has no equivalent framework
  • No oversight of frontier AI models – The most powerful AI systems face no mandatory safety testing or evaluation before release

78-93% of experts consider current government measures inadequate across five key AI threat categories, in Good Ancestors' AI Legislation Stress Test. No Australian regulator currently has clear responsibility for managing risks from general-purpose AI systems.

International comparison shows we're falling behind:

  • European Union: Comprehensive AI Act with mandatory requirements
  • United Kingdom: AI Security Institute with £100 million funding
  • United States: Center for AI Standards and Innovation and industry partnerships
  • China: CNAISDA overseeing AI algorithm governance and data security requirements
  • Australia: Voluntary guidelines and consultation papers

Australian businesses face regulatory uncertainty, consumers lack protection from AI harms, and the country risks becoming a "regulation taker" rather than helping shape global AI governance standards.

Not really. Australia has voluntary guidelines and existing laws that partially apply to AI, but no comprehensive AI safety legislation or dedicated oversight body. What we currently have: Voluntary AI Safety Standard – Guidelines that companies can choose to follow, with no enforcement mechanism or penalties for non-compliance Existing sector laws – Privacy Act, Consumer Law, and workplace safety rules that cover some AI uses, but weren't designed for modern AI systems Proposed mandatory guardrails – The government is consulting on rules for "high-risk" AI, but these remain undefined and unlegislated What we don't have: No AI Safety Institute – Unlike the US, UK, and other allies who have dedicated technical bodies to assess AI risks No comprehensive AI Act – The EU passed landmark AI legislation in 2024; Australia has no equivalent framework No oversight of frontier AI models – The most powerful AI systems face no mandatory safety testing or evaluation before release 78-93% of experts consider current government measures inadequate across five key AI threat categories, in Good Ancestors' AI Legislation Stress Test. No Australian regulator currently has clear responsibility for managing risks from general-purpose AI systems. International comparison shows we're falling behind: European Union: Comprehensive AI Act with mandatory requirements United Kingdom: AI Security Institute with £100 million funding United States: Center for AI Standards and Innovation and industry partnerships China: CNAISDA overseeing AI algorithm governance and data security requirements Australia: Voluntary guidelines and consultation papers Australian businesses face regulatory uncertainty, consumers lack protection from AI harms, and the country risks becoming a "regulation taker" rather than helping shape global AI governance standards.
View all frequently asked questions

Australians demand immediate AI safety action

Recent polling shows overwhelming public support for Australian Government action on AI regulation and safety policies.

94%

Australians expect AI safety standards at least as strict as commercial aviation

1 of 6
94%

Australians expect AI safety standards at least as strict as commercial aviation

83%

Australians think AI regulation is falling behind the technology

77%

Australians do not trust tech companies to ensure their AI is safe

About Australians for AI Safety

Australians for AI Safety is a coalition advocating for the safe and responsible development of artificial intelligence in Australia. We bring together experts in AI, ethics, policy and other fields with concerned citizens so that Australia leads in AI governance.

Our mission is to promote informed public discussion about AI risks and benefits, and to translate expert knowledge into accessible policy recommendations. We advocate for governance frameworks that protect Australian interests while supporting beneficial AI development.

Through open letters, expert testimony, media engagement and grassroots advocacy, we work to make sure Australia's voice is heard in global AI governance discussions, and that government policy keeps pace with the technology.