Global AI regulation demands mount after PHP flaws expose cyber risks

Organisations around the world have rushed to fix critical vulnerabilities in PHP, the server-side language that powers a vast proportion of the global web, following warnings that attackers could seize control of unpatched infrastructure.

The emergency has sharply refocused international attention on artificial intelligence. While researchers used AI tools to discover the flaws, automated systems also threaten to dramatically accelerate attempts to exploit them. Amazon confirmed it detected exploitation attempts even before the vulnerabilities were publicly disclosed.

The incident follows a series of automated intrusions that have put security teams on alert worldwide. Advanced models have recently been tied to breaches targeting an Australian health department platform as well as a automated bot attack on Hugging Face, an open-source machine learning hub.

Automated threats push governments toward guardrails

Security officials argue that the speed of AI-driven breaches makes national defences alone insufficient. In Britain, AI minister Kanishka Narayan warned that an automated attack on domestic infrastructure remains "likely, if not inevitable," pointing to the limits of purely local responses.

While the UK is advancing domestic legislation to shield digital networks and leaning on the work of its AI Security Institute in London, advocates for multilateral standards are pushing to place binding AI oversight on the international agenda. Plans have been mooted to elevate global AI governance during Britain's upcoming G20 presidency.

Yet diplomatic consensus remains deeply fractured. The United States administration under President Donald Trump has dismissed international regulatory negotiations, treating frontier AI capability as a direct contest for technological dominance against China. "Whoever wins AI, wins," Trump has maintained, resisting statutory curbs even as several technology executives call for external safety baselines.

The narrowing gap between the US and China

That race for supremacy is tightening faster than Western policymakers anticipated. Former Google chief executive Eric Schmidt warned that China is now trailing the United States by less than six months in AI development, down from an estimated lead of one to two years.

Speaking on the In Good Company podcast hosted by Nicolai Tangen of Norges Bank Investment Management, Schmidt argued that technological sophistication alone may not decide the contest.

Chinese developers have narrowed the distance through software innovations like AI distillation—training lightweight models on outputs from larger systems—alongside access to hardware despite export controls, Schmidt noted. By offering free or low-cost open-weight models, Chinese companies could secure rapid adoption across emerging markets.

Schmidt pointed to practical deployments such as Tencent's WeChat assistant, which reads medicine packaging to advise consumers, as proof that commercial reach often hinges on price rather than pure compute.

Industry analysts note that while the White House views safety regulations as an impediment to beating foreign rivals, software emergencies like the PHP scramble show that technological leads offer zero immunity against global digital contagion.