EU AI Act nears enforcement, SMEs warned
The EU’s landmark AI Act is moving toward enforcement for businesses in August 2026, but a European Parliament committee has backed delaying certain provisions until 2027 while keeping a fixed deadline—noncompliance can trigger fines up to €35 million or 7% of global turnover. (kinstellar.com) Regulators and experts say SMEs face a “dangerous blind spot” because routine software could fall under the rules, and the rollout raises equity questions (languages like Catalan risk exclusion); hundreds of protesters also targeted OpenAI and Anthropic this week calling for a slowdown in AI development. (xpert.digital) (en.ara.cat) (ppc.land) (indiatoday.in)
IMCO and LIBE adopted a joint draft report (Digital Omnibus on AI, reference A10-0073/2026) on 18–19 March 2026, with co-rapporteurs Arba Kokalari and Michael McNamara and a committee vote reported as 101 in favour, 9 against and 8 abstentions. (ppc.land)) The draft replaces the Commission’s conditional “trigger” approach (COM(2025)0836) with calendar deadlines and specifies application dates of 2 December 2027 for Annex III high‑risk systems and 2 August 2028 for Annex I systems. (ppc.land)) The Commission had originally proposed linking high‑risk obligations to the availability of harmonised standards and a six‑month countdown; the Parliament’s draft removes that link and argues delays in standards, guidelines and conformity frameworks justify fixed postponements. (euractiv.com)) SMEs are explicitly referenced throughout the Act (the text highlights SMEs multiple times), and industry groups such as the Digital SME Alliance and Eurochambres have warned that unclear boundaries risk capturing routine tools used by small firms—examples flagged in commentary include CV‑screening and other HR or administrative automation. (artificialintelligenceact.eu)) Article 99 of the AI Act lays out tiered penalties beyond the headline figure, including sanctions up to €35 million or 7% of global turnover for the gravest breaches, up to €15 million or 3% for other infringements, and up to €7.5 million or 1% for providing incorrect or misleading information. (artificialintelligenceact.eu)) Catalonia has promoted an AI assessment model adopted by the Basque Country, Croatia, Malta and Brazil as a regional approach to rights‑safeguarding under the Act, while Spain’s push to make Catalan an official EU language has repeatedly failed to secure unanimous support from member states. (en.ara.cat)) Organisers under the “Stop the AI Race” banner staged marches in San Francisco on 21 March 2026 that moved from Anthropic’s HQ to OpenAI and xAI, with Reuters and local outlets reporting dozens to hundreds of participants calling for CEOs to commit to a conditional pause on frontier AI development. (reutersconnect.com))
