The legal framework for the safe deployment of Artificial Intelligence Artificial Intelligence has arrived in companies – often faster than rules and responsibilities can keep pace. As of 2 August 2026, the market surveillance authorities take up their supervisory activities, and with the German AI Market Surveillance and Innovation Promotion Act (KI-MIG) it is now settled who inspects and imposes sanctions in Germany. At the same time, the Digital Omnibus postpones the obligations for high-risk AI. So what actually applies now – and what does it mean for occupational safety and health? This talk provides a compact, practice-oriented overview of the three dimensions of safe AI deployment: the legal obligations arising from the AI Act, the GDPR and the German Works Constitution Act; the new attack surfaces for information security; and – as the main focus – the consequences for occupational safety and health in the workplace. Many companies are still unaware that the AI literacy obligation has applied to every deployer since February 2025. At the centre stands the risk assessment under Section 5 of the German Occupational Safety and Health Act (ArbSchG): it is the instrument with which occupational safety specialists systematically capture AI-related risks – from psychological strain and automation bias, through the traceability of system decisions, to accountability in the event of system failure. Drawing on concrete examples from the daily work of the safety specialist, the talk shows where AI strengthens occupational safety and where it creates new hazards – for instance along the fine line between fatigue detection permitted for safety reasons and the emotion recognition in the workplace prohibited under Article 5 of the AI Act. Participants take away a seven-step roadmap for introducing AI in their own organisation in a way that is legally sound, cyber-secure and safe at work.