UN AI Panel Urges Global Governance as Costs Surge and New Low‑Cost Models Emerge

Italian Prime Minister Giorgia Meloni warned that AI‑generated texts, images and videos could undermine democratic reality unless the technology is made transparent and traceable. She stressed that “the base of common reality” is at risk and called for clear rules on AI content provenance.

The United Nations’ Independent International Scientific Panel on AI released its first global scientific report, warning that without coordinated rules the rapid advance of generative models could cause “catastrophic” damage. Co‑chairs Maria Ressa and Yoshua Bengio highlighted the gap between fast‑moving technology and slow‑moving policy, urging immediate international oversight.

Anthropic announced the launch of Sonnet 5, an agentic model that matches high‑end competitors at roughly one‑third the price, promising to democratise access to advanced AI capabilities.

Across the corporate sector, firms such as Citi, Atlassian, Adobe, Amazon and others are curbing spending on premium AI models after token‑usage costs spiked dramatically. Internal documents show limits being placed on models like Claude Opus 4.6/4.7 and GPT‑5.5, with a shift toward cheaper alternatives to control budgets that have risen by up to 500 % in six months.

Italian‑based Regolo.ai, built by Seeweb, promotes an open‑weight, multi‑provider approach to avoid lock‑in, offering models from Alibaba, Google DeepMind, Meta and others while keeping API compatibility with OpenAI.

At the AWS Summit in Milan, data showed that 40 % of Italian firms use AI, yet only a small fraction have integrated it beyond experimentation, highlighting a gap between adoption and transformative impact.