NEW YORK — What happened
Following the enforcement deadlines of the EU AI Act in mid-2026, the European AI Office has published its first public summary of compliance disclosures for high-risk systems. The regulatory filings focus heavily on automated recruitment, CV screening, and worker evaluation systems deployed across member states.
Why it matters
This marks the first real-world test of regulatory policy governing artificial intelligence at scale. The results offer concrete data on how far current commercial software falls short of legally mandated standards for fairness, transparency, and human oversight.
Deep dive
Of the initial batch of 350 audited enterprise HR systems, 42% failed to meet mandatory thresholds for statistical parity or post-hoc auditability. The primary bottleneck was not deliberate bias, but poor technical documentation and an inability to explain why specific candidate recommendations were generated. Vendors frequently relied on proprietary API calls to third-party model providers, creating a 'black box chain' that prevented thorough risk assessment.
Report check
The regulatory summary explicitly notes that three major HR software providers have received formal compliance warnings. This is verified by public regulatory enforcement dockets. Claims circulating in industry forums that massive multi-million-euro fines have already been issued are currently unverified and appear to misinterpret preliminary notices of inquiry.
Open questions
Will software vendors simplify their AI stacks to meet strict regulatory reporting requirements, or will high-risk applications migrate outside the EU jurisdiction? How will small businesses balance the cost of independent algorithmic audits?