EU AI Act Forces New Deepfake Labels and Triggers Broad Political Debate on AI Governance

From August 2 2026 the European Union’s AI Act will require all synthetic audio‑video content to carry a digital watermark and a clear label for users, a rule aimed at stopping deepfake fraud and protecting personal dignity. The regulation, described by EU officials as a “war on false AI‑generated content”, has met resistance from Meta, which has rejected the agreement.

In Italy, Eurodeputy Alberico Gambino warned that regulation alone is insufficient, urging Europe to build its own semiconductor, data‑center and critical‑materials supply chains to secure strategic autonomy. The same week the regional council of Tuscany approved a law creating an AI observatory, digital‑facilitation points and worker‑retraining programmes, while Italy’s communications regulator Agcom announced a new AI office to safeguard pluralism and enforce human oversight of algorithms.

Parallel to the policy push, a Stanford‑led open letter signed by more than 200 scholars and 16 Nobel laureates called for urgent rules to manage AI’s economic impact, warning that AI‑linked layoffs in the United States hit a record 38,579 jobs in May 2026 – 40 % of all cuts, affecting roughly 150 000 workers. A McKinsey report highlighted that by 2026 cloud giants will invest over $700 billion in AI infrastructure, while Google launched a series of free one‑hour AI courses. Researchers such as Ines El Gataa warned that AI systems reflect the biases of the data and the dominance of U.S. firms like Nvidia, reinforcing a “compute divide”.

These developments were discussed at the European “Artificial Intelligence for Accessibility” conference in Brussels, where an avatar capable of signing in Italian Sign Language was demonstrated, showcasing AI’s potential for inclusive services.