EU regulators launch inquiry into X’s Grok AI as usage surges 22% following European expansion, examining risks of its unfiltered ‘Unhinged’ mode that prioritizes raw public data over conventional safeguards.
European regulators have opened an investigation into X’s Grok AI as its controversial ‘Unhinged’ mode gains traction across 15 new EU countries, amplifying debates about AI ethics.
The European Union’s AI regulatory body announced a formal inquiry this week into X’s Grok artificial intelligence system, specifically targeting its ‘Unhinged’ mode that delivers intentionally unfiltered responses. This comes just days after X expanded Grok’s availability to 15 additional European countries on 18 July 2025 despite ongoing concerns about content moderation protocols.
Controversial AI Approach
Grok’s signature feature trains exclusively on public conversations from X’s platform, deliberately bypassing conventional content filters. Elon Musk has championed this approach as ‘truthful AI’ that reveals what competitors withhold, telling investors last month: ‘If sanitized AI is aspirin, Grok is the full chemical spectrum.’ Recent Stanford University research published on 17 July appears to validate concerns, showing Grok’s unfiltered outputs generate controversial statements three times more frequently than OpenAI’s GPT-4.5.
Regulatory Pushback
EU regulators are examining whether Grok’s architecture violates Article 5 of the AI Act prohibiting systems that deploy subliminal techniques or exploit vulnerabilities. ‘When platforms intentionally disable risk mitigation layers, they potentially endanger users,’ stated Margrethe Vestager, Executive Vice-President of the European Commission for A Europe Fit for the Digital Age. The inquiry coincides with X’s own data showing Grok now processes 45% more real-time platform data than in Q1 2025, as Musk revealed in a 20 July tweet.
Market Impact
Despite regulatory challenges, analytics firm Similarweb reports Grok usage surged 22% since June 2025, with adoption strongest among 18-34 year-olds. AI Now Institute researchers warn this growth may come at significant societal cost. ‘Unfiltered training on public posts replicates existing platform biases at scale,’ said Director Amba Kak, noting that X’s own 2024 transparency report documented a 33% increase in hate speech reports prior to Grok’s expansion.
The current regulatory confrontation echoes Microsoft’s 2016 shutdown of its Tay chatbot within 24 hours after users manipulated it into producing racist and sexually charged content. Like Grok, Tay learned from unfiltered public interactions but lacked adequate guardrails. That incident became a case study in AI ethics courses worldwide and contributed to today’s regulatory frameworks.
Historically, the tension between raw data access and ethical constraints traces back to early web crawlers in the 1990s. Stanford’s original PageRank algorithm similarly leveraged public link structures but established quality filters that became industry standards. Grok’s approach represents a philosophical departure from this tradition, prioritizing immediacy over curation in ways that both challenge conventional AI development and test regulatory boundaries.
