# Beijing reclassifies artificial intelligence as a security problem as Canberra presses Apple on online safety

China's state security ministry warns that the hacking capability of AI systems endangers national security, while Australia's prime minister asks Tim Cook to back its internet rules.

- Published: 2026-09-20T05:12:09.557Z
- Canonical: https://polylog.news/2026-09-20/beijing-reclassifies-artificial-intelligence-as-a-security-p
- Publisher: Polylog (Global desk)
- Section: tech
- Sources: [The Japan Times](https://www.japantimes.co.jp/news/2026/09/20/asia-pacific/politics/china-ai-security-risks/), [Al Jazeera](https://www.aljazeera.com/news/2026/9/20/australia-seeks-big-tech-support-for-internet-safety-ai-regulation)

Governments on three continents spent the weekend treating artificial intelligence as a liability to be contained rather than a growth programme to be accelerated.

The Japan Times reports that the hacking ability of AI systems has led Beijing to view the technology through a security lens, with the Ministry of State Security warning that [national security could be endangered](https://www.japantimes.co.jp/news/2026/09/20/asia-pacific/politics/china-ai-security-risks/). That is a notable shift for a state that has made AI central to its industrial strategy and to its answer to United States export controls on advanced semiconductors. A security ministry that treats domestically developed models as an attack surface (a potential entry point for hackers) will push for approval requirements and access restrictions that slow the same deployment the industrial policy is trying to speed up.

In Australia, Prime Minister Anthony Albanese asked Apple chief executive Tim Cook to support Canberra's online safety laws, [Al Jazeera reported](https://www.aljazeera.com/news/2026/9/20/australia-seeks-big-tech-support-for-internet-safety-ai-regulation), as the government extends its regulatory agenda from social media age limits toward AI. Australia is small enough that companies could ignore it, and it has instead become a test case for whether device makers and platforms will implement age and safety controls at the operating system level.

The liability question is arriving through courts as well. The South China Morning Post reports that a man in Zhejiang province [sued an AI company](https://www.scmp.com/news/people-culture/trending-china/article/3368136/chinese-man-sues-ai-firm-after-chatbots-auspicious-date-suggestion-leads-disaster?utm_source=rss_feed) after following a chatbot's recommendation for his mother's burial date and blaming subsequent family misfortune on the advice. The claim is unusual, but the principle it tests is not. Courts will have to decide when a model's output constitutes advice for which its maker is answerable.

Each of these pressures adds cost to the same business model. Security review, age verification and legal exposure all reduce the margin that investors have been extrapolating from AI deployment.

## What this means

Equity valuations in AI infrastructure and platform companies rest on rapid, low-friction deployment, and security clearance requirements, safety mandates and product-liability claims each insert delay and cost between capital spending and revenue. Chinese model developers face the sharpest version, because a state security ministry warning can restrict domestic rollout at the same time export controls limit the chips available to build with.

## What to watch

- Whether China's Ministry of State Security or the Cyberspace Administration issues formal rules rather than warnings, which would set an actual compliance cost for Chinese model developers.
- Apple's and other device makers' response to Australia, since agreement there establishes a template other regulators will copy.
- Capital expenditure guidance from the largest AI infrastructure buyers, the clearest signal of whether regulatory friction is changing spending plans.