Dividing AI into 'virtuous' West, 'sinister' others risky algorithm

Artificial intelligence seems to have acquired a nationality, a political conscience and a price tag for "virtue" to some.

In a recent article in The Guardian, the author warns the United Kingdom against cheap Chinese AI, suggesting that the United States' technology comes with safeguards and benign intentions that the Chinese alternatives lack. Unfortunately, the argument rests comfortably on Cold War prejudice.

There are legitimate concerns about AI's safety, but that concern should be based on technological evidence. Recent disclosures offer a rather inconvenient rebuttal to the notion of inherently "safer" US AI.

During cybersecurity evaluations, OpenAI models, including an internal research model tested alongside GPT-5.6 Sol, circumvented sandbox restrictions, exploited vulnerabilities and compromised parts of Hugging Face's production infrastructure. Anthropic's Claude submitted a fabricated homicide tip to Philadelphia police and, in separate tests, interacted with government websites in unintended ways. Google's Gemini accessed systems belonging to three real companies during security testing, while Meta disclosed that its Muse Spark model exploited a vulnerability in a third-party service. Microsoft's Copilot, meanwhile, faced a reported data-exfiltration vulnerability known as CoSnitch.

These incidents differed in cause and consequence, but they expose a common weakness: sophisticated safeguards and US corporate credentials are no guarantee of security. Silicon Valley, it turns out, has no monopoly on technological virtue.

The Guardian article baselessly portrays non-Western AI models as instruments of "control" while overlooking the extraordinary technological machinery already available to Western governments for that very purpose.

In the UK and the US, for instance, the practices go even further but are defended as "necessary" for national security, crime prevention and administrative efficiency. But their surveillance does not become harmless simply because it is conducted under a so-called "democratic" flag.

Indeed, the distinction between open-source and closed-source AI deserves more attention than the article gives it. Chinese developers distribute models that organizations can download, inspect, modify and operate locally. This allows users to retain greater control over sensitive information rather than automatically transmitting it to a commercial provider.

Closed-source AI presents its own political risks. When Western governments purchase opaque systems from powerful technology companies, commercial secrecy can obstruct scrutiny, while long-term contracts encourage dependence on a handful of suppliers.

Who can guarantee that for-profit companies won't do evil? The danger is not hypothetical. US immigration enforcement increasingly relies on privately supplied data-analysis and biometric technologies. The resulting concentration of information and decision-making power raises uncomfortable questions.

There is also an economic argument that The Guardian article's geopolitical anxieties obscure. Open-source AI models challenge a US business model built around expensive computing infrastructure, proprietary access and recurring subscriptions. As investors question whether enormous AI spending will generate adequate returns, cheaper competitors threaten more than technological prestige: they threaten profitability.

Perhaps this explains some of the eagerness to equate affordability with "insecurity". If consumers can obtain capable technology at a fraction of the price, the premium charged by established suppliers requires justification beyond national origin.

The UK should certainly scrutinize AI. But that scrutiny should not be discriminatory. A safety challenge does not become less serious just because it is caused by a Western AI model.

The most dangerous algorithm may be the intellectual shortcut that divides technology into "virtuous" Western technology and "sinister" others. That is ideology masquerading as consumer protection.

When the arrogance underpinned by a Western-centric approach faces a real-world challenge from developing nations, the act of smearing the latter as "cheaply evil" while lauding the former as "expensively virtuous", in itself, exposes anxiety rather than insight.