AI skills and governance in focus as models advance

The rapid pace of development in frontier artificial intelligence models is driving renewed debate about governance, security and workforce capability, as organisations and policymakers respond to evolving cybersecurity risks and growing competition between AI developers.

Recent developments surrounding access restrictions on Anthropic's Claude Mythos model have highlighted how governments are weighing national security concerns against the commercial and strategic importance of advanced AI systems. At the same time, Australia's technology sector is calling for greater investment in AI skills alongside stronger governance frameworks.

Model access

Heather Ceylan, Chief Information Security Officer, Box, said recent events demonstrate how quickly AI model availability can become a policy issue extending beyond the technology industry.

"The debate around Anthropic's Claude Mythos has already moved beyond Silicon Valley. Recently, access to the model became the subject of unprecedented government restrictions. This is primarily over concerns that its vulnerability-discovery capabilities could be misused by hostile actors. Controls have partially eased following additional safeguarding," said Heather Ceylan, Chief Information Security Officer, Box.

The temporary restrictions have drawn attention to the increasing role advanced AI models play in cybersecurity. Models capable of identifying software vulnerabilities can assist defenders in finding and addressing weaknesses more quickly. The same capabilities also raise concerns that malicious actors could use them to improve offensive cyber operations.

Ceylan said competition between AI developers is also reshaping the market as organisations gain access to a broader range of foundation models.

"In a matter of weeks, this has created an interesting ripple effect: AI supply is becoming increasingly global. Foundational models are advancing at extraordinary speed, with capabilities improving in a matter of weeks rather than years. There is now a low tolerance for interruptions or restrictions to AI model access. Alongside the leading US developers, Chinese AI models are gaining ground with enterprises around the world, including in the US. This is being driven by lower costs, rapidly improved performance and growing model choice. As the market matures and becomes more competitive, organisations are increasingly selecting models based on capability and business value, rather than country of origin alone," said Ceylan.

Governance focus

As organisations evaluate AI platforms from a growing number of providers, governance requirements are becoming a larger consideration alongside model performance and cost.

Enterprises are increasingly examining issues such as privacy, security, regulatory compliance, transparency and auditability before deploying AI technologies across business operations.

Ceylan said the changing competitive landscape reinforces the need for stronger governance as AI capabilities continue to develop.

"That makes governance more important than ever. Every non-native AI system, developed outside an existing technology ecosystem and legislature, needs to be subjected to rigorous scrutiny around security, privacy, compliance, transparency and auditability. The temporary restrictions on Mythos ultimately underscored a broader reality, which is that frontier AI models are becoming powerful cybersecurity assets. They can help defenders identify and remediate vulnerabilities at unprecedented speed, but they can also be exploited by malicious actors. As AI reshapes cybersecurity, organisations must ensure they can harness these capabilities responsibly and faster than attackers," said Ceylan.

Skills investment

The Australia Computer Society has also called for increased investment in AI capability, arguing that workforce development will be central to Australia's ability to benefit from AI technologies.

The organisation said AI policy should extend beyond regulation and infrastructure to include sustained investment in skills development across the workforce.

"Ultimately, AI will be shaped by the people who build it, govern it and use it. If we want Australia to lead, we have to treat AI and cyber security skills as core national infrastructure - that means skilling the whole workforce and deliberately lifting the capability of our professionals in AI and cyber, not leaving it to chance.

"Australia now has to do the hard work: identify the gaps, fund accredited micro‐credentials at scale, and equip workers to lead change rather than simply absorb it. If the Prime Minister is serious about an AI-enabled economy, then the test will be whether government backs workers with real investment in AI skills and training, so we lift productivity and create the future here, instead of importing it from somewhere else," said Dr Prins Ralston, Chief Executive Officer, Australia Computer Society.

The society also welcomed efforts to involve workers in discussions about the future use of AI across Australian workplaces.

"That responsibility cannot be left to individuals or employers to navigate alone. ACS welcomes the Fair AI Taskforce and its commitment to giving workers a greater say in how AI shapes Australia.

"We look forward to contributing our expertise in workforce capability, professional standards and trusted skills development," said Beau Tydd, President, Australia Computer Society.