Alexandru Voica, Head of Corporate Affairs & Policy at Synthesia, navigates the complex intersection of AI regulation and startup innovation, drawing on 15 years of tech experience across ARM and Meta to educate policymakers about artificial intelligence while protecting European startups from regulatory overreach. Romanian-born Alexandru advocates for startups in EU AI Act discussions, highlighting how fine-tuning requirements can burden small companies with full model provider obligations when modifying open-weight models like Llama for specialised healthcare applications, forcing them to produce documentation, training data summaries, and copyright policies that larger companies can handle but startups struggle to manage. He critiques the EU AI Act's "Christmas tree" evolution where competing priorities from different member states and companies created unwieldy legislation with late-arriving implementation details like codes of practice, potentially slowing European innovation while US and Chinese competitors advance at "three to four times the speed" without such regulatory complexity. Alexandru questions the Act's broad "high-risk" classifications, using personal examples of learning English in Romania without wealthy family resources to illustrate how AI language learning tools could democratize education access, yet face the same regulatory burden as biochemical weapons under current frameworks. His policy work involves constant education of government officials who predominantly come from legal or Oxford PPE backgrounds rather than engineering, contrasting unfavorably with the Chinese leadership's engineering expertise that enables practical technology understanding and faster innovation cycles. At Synthesia, Alexandru advocates moving corporate communications away from traditional marketing-led PR agency relationships toward in-house corporate affairs functions that work closely with legal teams, emphasising quality press coverage over quantity to build long-term relationships and protect executive reputations. He champions "common sense" as his most valuable skill, sharing examples of preventing reputational disasters like engineers presenting monkey-snorting-data memes at major tech conferences, while expressing scepticism about AI-first companies like Cluely that sacrifice human relationships for viral attention, arguing that sustainable customer relationships require maintaining human touchpoints even within AI-automated systems.