Welcome to Daily Inference – your daily AI briefing from dailyinference.com. Today: The EU just launched a formal investigation into Google's AI training practices, raising questions that could reshape how tech giants build their models. Meanwhile, Moonpig is proving that AI isn't just hype – it's actually driving sales growth. And researchers are getting called out for the quality crisis in AI-generated content.

Brussels just escalated the AI regulation game. Yesterday, the European Union launched a formal investigation into Google's use of online content to train its AI models, specifically targeting how the tech giant scrapes and utilizes web data for its Gemini AI platform. This isn't a routine inquiry – it's a full-scale probe into whether Google's data collection practices comply with EU digital regulations.

The investigation centers on transparency and consent – two pillars of European data protection law. EU regulators want to know exactly how Google collects training data from websites, whether content creators have meaningful ways to opt out, and if the company properly discloses its data scraping activities. The probe comes as AI companies increasingly face scrutiny over training their models on copyrighted material, news articles, and creative works without explicit permission or compensation.

This investigation could set precedent for the entire AI industry. If the EU finds violations, Google could face substantial fines and be forced to fundamentally alter how it trains Gemini and future models. More importantly, any ruling will likely influence how other AI companies – from OpenAI to Anthropic – approach training data in European markets. For website owners and content creators, this probe represents a potential shift in power dynamics, possibly giving them more control over whether their work feeds AI systems.

While most companies talk about AI strategy, UK-based greeting card company Moonpig is actually making money from it. The online card retailer reported yesterday that its AI-powered design and personalization features are directly driving sales growth, offering a refreshing case study of generative AI delivering measurable business results rather than just generating hype.

Moonpig has integrated AI into its card creation process, allowing customers to generate personalized designs and customize messages with greater ease and creativity. The AI helps users overcome the blank-page problem that often stalls greeting card purchases, suggesting designs, improving message wording, and enabling customization that would typically require graphic design skills. This practical application addresses a real customer pain point – the difficulty of creating something personal and visually appealing quickly.

The results speak to a broader lesson about AI adoption: success comes from solving specific problems, not chasing technology trends. Moonpig didn't rebuild their entire business around AI or launch a flashy chatbot. Instead, they identified friction points in the customer journey and applied AI precisely where it added value. For businesses considering AI investments, Moonpig's approach offers a blueprint – find concrete use cases where AI reduces customer effort or enhances experience, implement thoughtfully, and measure actual business impact rather than vanity metrics.

Speaking of practical AI applications, if you're looking to quickly build a professional web presence, check out 60sec.site – an AI-powered website builder that demonstrates how AI can simplify complex tasks without requiring technical expertise.

The AI research community is facing uncomfortable criticism about content quality. A letter published yesterday argues that AI researchers themselves bear responsibility for the flood of low-quality AI-generated content – often called 'slop' – that's degrading online information ecosystems. The critique strikes at a sensitive nerve: the people building AI systems may be inadvertently enabling their misuse.

The argument centers on responsibility and foresight. The letter suggests that researchers have prioritized technical capabilities and benchmark performance while insufficiently considering how their models would be deployed at scale. As generative AI tools became accessible to millions, they've been used to mass-produce mediocre content for SEO spam, social media manipulation, and academic fraud. The critics argue that researchers can't simply build powerful tools and then disclaim responsibility for their societal effects.

This criticism reflects a broader reckoning in the AI field about ethical responsibility and deployment consequences. Unlike previous waves of technology where misuse emerged gradually, AI-generated content pollution happened rapidly and at scale. The debate raises fundamental questions: Should researchers build in stronger safeguards before release? Do they have obligations beyond technical innovation? How can the field balance open research with preventing harm? As AI capabilities continue advancing, these questions become increasingly urgent – and the research community will need better answers than 'we just build the technology.'

Today's stories reveal three parallel tracks in AI's evolution. Regulators are finally catching up with enforcement actions that could reshape industry practices. Meanwhile, practical applications like Moonpig's show that AI can deliver real business value when thoughtfully applied. And the growing criticism of content quality reminds us that technical capability doesn't equal responsible deployment. The tension between innovation speed and societal impact remains AI's defining challenge.

Stay informed with daily AI insights at dailyinference.com. We cut through the hype to bring you what actually matters in artificial intelligence.