EU AI Act takes effect as UK survey reveals deep public scepticism over AI-generated content

The European Union's Artificial Intelligence Act, the world's first comprehensive legal framework for AI, entered into force on Monday 3 August after its adoption in 2024. The legislation introduces a tiered risk-based approach that bans certain applications outright, imposes strict obligations on high-risk systems, and requires transparency for generative AI, including the labelling of synthetic content. Organisations that breach the rules face fines of up to €35 million or 7 percent of global annual turnover, whichever is higher.

While the Act is an EU instrument, its extraterritorial reach means any company offering AI services in the single market must comply, regardless of where it is based. The United Kingdom, having left the EU, is not bound by the Act directly, but British firms serving European customers fall within its scope. The UK government has meanwhile pursued its own, lighter-touch approach to AI governance, relying on existing regulators and voluntary commitments rather than a standalone statute.

What the Act actually requires

The regulation classifies AI systems into four risk categories. Unacceptable-risk systems, such as social scoring by governments or real-time biometric identification in public spaces, are prohibited. High-risk systems, covering areas from medical devices to recruitment tools, must undergo conformity assessments, maintain detailed technical documentation, and register in an EU database. Limited-risk systems, which include most generative AI models, face transparency obligations: users must be informed they are interacting with AI, and synthetic audio, video, text or images must be marked as artificially generated. Minimal-risk applications, such as spam filters or video-game AI, are largely unregulated.

The transparency rules for generative AI are where the Act intersects most directly with public concerns about content. Providers of general-purpose AI models, the large language models that power chatbots, image generators and coding assistants, must publish summaries of training data, respect copyright law, and ensure downstream users can comply with labelling requirements. The European Commission has said it will publish codes of practice to guide compliance, but these are not yet finalised.

British public opinion: labelling is non-negotiable

Against this regulatory backdrop, YouGov surveyed British adults to gauge attitudes toward AI-generated content. The headline finding is unambiguous: 89 percent of respondents say it is important that content explicitly states when it has been created by AI. Only a small minority consider labelling unimportant. The figure suggests a near-consensus on transparency that cuts across demographic lines, and it aligns closely with the Act's own disclosure requirements.

The same survey finds that 70 percent of UK adults agree with the statement that "there may not be enough rules or oversight for how AI is used." This is a striking number given that the EU has just enacted the world's most ambitious AI rulebook. It points to a gap between legislative action and public perception, or perhaps a scepticism that rules on paper will translate into effective enforcement. The Act's phased implementation, with most provisions not applying until 2026 and 2027, may also contribute to the sense that oversight remains theoretical.

Trust, deception and the brand risk

YouGov asked respondents which scenarios would cause them to lose trust in a brand using AI-generated content. The top answer, selected by 70 percent, was content that "feels misleading or deceptive." Close behind, at 68 percent, was content that does not disclose AI usage. These two concerns are effectively two sides of the same coin: non-disclosure is experienced as deception. For brands, the message is clear, transparency is not a regulatory box-ticking exercise but a commercial necessity.

The data also reveals a nuanced hierarchy of discomfort. More than half (56 percent) would lose trust if AI replaces humans entirely in a given function. This suggests the public distinguishes between AI as a tool that assists human creators and AI as a substitute that eliminates them. The distinction matters for sectors such as customer service, journalism and creative industries, where the economics of full automation are tempting but the reputational risk is now quantified.

Sensitive domains: news, health and politics

Nearly half of respondents (49 percent) say they would lose trust if AI-generated content is used across sensitive topics such as news, health or politics. This finding resonates with broader debates about synthetic media in elections, AI-written medical advice, and automated news aggregation. The EU Act does not ban AI in these domains, but it does classify certain applications, such as AI systems used to influence voting behaviour or provide medical diagnoses, as high-risk, triggering the strictest obligations.

The sensitivity finding also highlights a tension for platforms and publishers. News organisations are experimenting with AI for summarisation, translation and even drafting routine stories. Health providers are trialling chatbots for triage. Political campaigns are testing synthetic media for outreach. Each use case now carries a measurable trust penalty if disclosed poorly or not at all.

A positive minority: where the public sees value

The survey is not uniformly negative. Forty-one percent of UK adults agree that generative AI can help with "generating ideas when people are stuck," and 38 percent see value in "content creation more generally." These figures represent a substantial minority that views AI as a creative aid rather than a threat. The gap between the 89 percent demanding transparency and the 38, 41 percent seeing creative utility suggests the public wants AI to be visible and accountable, not absent.

This duality, high demand for labelling alongside recognition of creative benefits, complicates the narrative of simple technophobia. It also mirrors the EU Act's own philosophy: the legislation does not ban generative AI, but it insists on transparency and accountability. Whether the public perceives the Act as delivering on that promise will shape the political sustainability of the framework.

Enforcement: the test the Act has not yet faced

The Act's penalties are severe on paper. A €35 million fine or 7 percent of global turnover exceeds the maximum sanctions under the General Data Protection Regulation (GDPR), which caps at €20 million or 4 percent. But GDPR enforcement was slow to ramp up; the first major fines arrived years after the regulation took effect. The AI Act creates a new governance structure, national market surveillance authorities coordinated by a European AI Office within the Commission, but staffing and expertise remain open questions.

The Commission has signalled that enforcement will initially focus on prohibited practices and high-risk systems placed on the market without conformity assessment. Generative AI transparency obligations, while legally binding, may see a softer start as codes of practice are finalised. This phased approach is pragmatic but risks reinforcing the public perception that oversight is inadequate, the very concern the YouGov data captures.

Implications for companies operating in Europe

For businesses, the combination of regulatory risk and reputational exposure creates a dual imperative. Compliance with the Act's labelling and transparency rules is now a legal baseline. But the survey data suggests that merely meeting the legal minimum, a small "AI-generated" watermark, for instance, may not satisfy the 70 percent who react to content that "feels misleading." The standard is perceptual, not just procedural.

UK-based firms have an additional calculation. They are not subject to the Act domestically, but they are subject to it for any EU-facing service. The UK's own regulatory direction, currently guided by the Department for Science, Innovation and Technology and existing regulators such as the Information Commissioner's Office and Ofcom, may diverge. A dual compliance burden is emerging, and the YouGov data suggests British consumers would welcome stronger domestic rules, not weaker ones.