Reece Appleton, regional director for APAC at Huntress, said the push for early access to advanced AI systems should extend beyond government bodies so that local security researchers and frontline cyber teams can test how the technology might be abused and develop detections before malicious actors deploy it in the wild.
The appeal comes amid a broader policy debate over AI guardrails and fresh reports that Google’s Gemini model inadvertently accessed three companies during a testing phase, highlighting the speed at which AI can automate familiar attack steps.
Appleton argued that the real danger is not autonomous machine intent but the way frontier AI lowers the skill threshold for common cyber‑crime. He warned that less‑experienced attackers could use the models to automate reconnaissance, craft convincing localized phishing messages and generate working exploits far more quickly than before.
Justin Allen, head of Huntress’s Security Operations Centre for APAC, clarified that Google’s test did not reveal a new class of attack. Instead, the AI chained together known techniques—identifying exposed credentials and guessing passwords—at a much faster pace, demonstrating how AI can compress weeks of manual work into minutes.
Both executives say policymakers must balance tighter oversight with mechanisms that allow vetted researchers to experiment with these models under controlled conditions. Without such access, Australian defenders risk lagging behind attackers, especially as small and mid‑market businesses—often lacking dedicated security staff—become more attractive targets when AI reduces the cost of probing them.
The firm stresses that core defensive measures such as multi‑factor authentication, least‑privilege access, patching and defence‑in‑depth remain essential, but organisations also need to invest in rapid detection and incident‑response capabilities to limit damage when automated attacks succeed.