AI Safety & Risk: What You Need to Know Right Now
By Saghir Hussain ·
The CMA’s £1bn Fine Is a Warning Shot, Not a Solution
Let’s cut the sci-fi crap. The biggest AI risk right now isn’t a rogue robot; it’s the unregulated, profit-driven systems already operating in our daily lives, and the UK’s competition watchdog just proved it’s not fit to stop them. The Competition and Markets Authority (CMA) has finally issued a provisional finding against Google for allegedly using its dominance in the digital advertising market to favour its own services. The proposed fine? A staggering £1.35 billion. This isn’t a minor regulatory slap on the wrist; it’s a landmark case that exposes the fundamental problem: our regulatory framework is built for a 20th-century industrial economy, not a 21st-century algorithmic one.
The CMA’s own investigation highlights how Google’s practices harm competition and, by extension, consumers and businesses. But here’s the critical point: the CMA is investigating behaviour, not the technology itself. It’s looking at the chessboard, not the pieces. The real risk lies in the AI systems that power these behaviours—the algorithms that decide what you see, what you pay, and who gets a fair chance. These are opaque, complex, and often unaccountable. The CMA can fine Google for anti-competitive conduct, but what happens when an AI, trained on biased data, systematically excludes certain UK small businesses from advertising platforms? Who is held responsible? The developer? The user who prompted it? The regulator is left grasping at shadows.
NHS Data and the Privacy Time Bomb
While the CMA battles in the commercial arena, a more personal risk is unfolding in our public services. The National Health Service (NHS) has been exploring AI for diagnostics and administrative efficiency, a noble goal. But the journey there has been rocky, revealing a cavalier attitude towards our data. We’ve seen reports of NHS Trusts sharing patient data with private tech companies without explicit, informed consent, under the guise of "improving outcomes." This is the real, present danger: the erosion of trust in a foundational public institution.
The risk isn’t a dystopian database hack (though that’s a concern); it’s the slow, quiet normalisation of our most sensitive health data becoming a commodity for training commercial AI models. When an AI makes a diagnostic error, the accountability chain vanishes. Is it the fault of the algorithm, the data it was fed, or the NHS Trust that deployed it? The current system offers no clear answers, leaving ordinary people to bear the consequences of decisions made by black-box systems they can’t scrutinise. This isn’t a future threat; it’s happening now, with our data, in our hospitals, and the UK government’s approach has been dangerously permissive, prioritising innovation over robust, independent oversight.
The Global Governance Gap and the UK’s Weak Position
The UK is posturing for a leadership role in "safe" AI, hosting a global summit and talking about a "pro-innovation" regulatory framework. But let’s be blunt: this is mostly hot air. The proposed framework relies on existing regulators like the CMA and the Information Commissioner’s Office (ICO) to police AI, using principles rather than hard, enforceable laws. This is a recipe for inaction. The ICO has been underfunded for years, and the CMA is clearly out of its depth with the technical complexities of machine learning.
Meanwhile, the real power lies with the US and China, where the largest AI models are developed with minimal regulatory interference. The UK has no sovereign cloud infrastructure to host sensitive data, no major AI research labs on the scale of Google DeepMind (which, ironically, is British-born but now an American asset), and limited leverage. We are essentially a rule-taker, not a rule-maker. Our "pro-innovation" stance risks becoming a race to the bottom, where we offer lax standards to attract tech investment, ultimately selling out the privacy and rights of our citizens for short-term gains. The critical question isn’t how the UK can lead in AI safety; it’s whether we have the political will to stand up to the tech giants and protect our own people, or if we’ll just become a compliant market for their unvetted products.