BRUSSELS — The European Parliament on Tuesday voted 412 to 189 to formally adopt the long-anticipated AI Liability Directive, completing a regulatory framework that will fundamentally reshape how artificial intelligence products are developed and sold across the 27-member bloc. The directive, which complements the EU's AI Act that came into force in 2024, establishes a clear burden-of-proof standard: companies deploying high-risk AI systems must proactively demonstrate those systems are safe, rather than waiting for consumers or regulators to prove harm occurred.

The vote marks the culmination of nearly two years of fierce lobbying from Silicon Valley firms and European tech startups alike, who argued that overly strict liability rules would drive innovation outside the continent. Under the final text, companies including Google, Microsoft, Meta, and Mistral AI face mandatory pre-deployment safety audits conducted by accredited third-party laboratories, with results submitted to the newly empowered European AI Office in Brussels. Products that fail or skip the process can be banned from EU markets within 30 days of a formal complaint.

European Commissioner for Digital Affairs Henna Virkkunen, who championed the directive's final revision, called Tuesday's outcome a 'historic moment for citizens' rights in the digital age.' Speaking at a press conference in the Hemicycle following the vote, she said the rules send a clear signal that 'the era of deploy-first, apologise-later is over in Europe.' Industry groups, including DigitalEurope, acknowledged the result while pledging to work with Brussels on implementation timelines they describe as 'technically aggressive.'

The directive's passage lands at a moment of heightened scrutiny over AI's societal impact. A survey released earlier this week by the European Consumer Organisation found that 68 percent of EU citizens support stricter government oversight of AI tools, particularly those used in hiring, credit scoring, and healthcare triage — precisely the high-risk categories targeted by Tuesday's rules. Critics from civil liberties groups, however, warn that the directive's enforcement mechanisms still lack teeth against non-EU-based model developers who license technology into Europe indirectly.

Markets responded swiftly. Shares in Paris-listed Mistral AI slipped 3.4 percent in afternoon trading as investors weighed compliance costs, while larger incumbents like SAP and Ericsson, which have built compliance infrastructure in anticipation of the rules, saw modest gains. Analysts at Bernstein Research wrote in a note that the directive 'effectively raises the cost of entry for smaller AI challengers in Europe, entrenching the position of well-resourced incumbents.' Full enforcement is set to begin on January 1, 2028, giving the industry approximately 21 months to align operations with the new framework.