A hiring manager does not need to call a tool artificial intelligence for the EU AI Act to notice it. If software ranks applicants, screens CVs, supports interview scoring, or feeds a promotion decision, the calendar now matters as much as the model card. Since 2 August 2026, the useful question in Germany is no longer whether the demo looked clever. It is whether the file exists. The file is not ceremonial. It is the place where HR, procurement, legal, and the vendor prove who uses the system, what it does, who supervises it, and which obligations were handled before candidates became test data. Every privacy lawyer has seen this movie: compliance that lives only in a slide deck tends to age badly. ## The date turns hiring AI into an operating condition EDGE Certified Foundation says the EU AI Act was adopted in 2024 and becomes generally applicable from 2 August 2026. JobCannon identifies the law as EU AI Act Regulation 2024/1689 and says AI systems used in employment, recruitment, and worker management are classified as high risk under Annex III section 4. That is the compliance trigger for German employers using automated hiring tools, and it is not improved by calling the system an assistant. JobCannon also places the AI Act inside Germany's existing employment and privacy stack, not above it in a neat cloud of policy optimism. Its Germany guide lists the Works Constitution Act from 1972, GDPR Article 22 from 2018, and the EU AI Act from 2024 as relevant rules for recruitment, assessment, and AI hiring tools. Translation: the new AI file must sit beside the works council file and the privacy file. If those are owned by three different teams, congratulations, you have found your first governance problem. ## Who is in scope is broader than the resume sorter EDGE Certified Foundation describes AI systems as increasingly shaping recruitment, performance management, promotion, workforce analytics, monitoring, and rewards. That framing matters because many employers still treat recruitment AI as a narrow CV parser issue. The risk classification is not limited to the first screening step if the system influences who is hired, who advances, or how workers are evaluated. JobCannon's Germany guide is blunt on the classification point: AI systems used in employment, recruitment, and worker management fall into the high risk category under Annex III section 4. High risk is not a ban, and it is not a press release adjective. It is a compliance status that says the system may be used only with the required controls, documentation, oversight, and registration work in place. ## What the law asks for in practice According to JobCannon, employers using AI hiring tools must conduct conformity assessments, maintain technical documentation, ensure human oversight, and register systems in the EU database. Put plainly, procurement needs more than a warranty that the vendor is responsible. A usable contract should say who supplies technical documentation, who supports the conformity assessment, how human oversight works in the product, and who handles the database registration step. This is where the usual corporate phrase about welcoming regulator guidance deserves a small eye roll. The obligation is not to welcome clarity. The obligation is to produce evidence. HR teams should be able to show when the tool is used, what decision it supports, what humans can override, and what records demonstrate the system was assessed before deployment. ## Germany adds works councils and privacy to the queue JobCannon notes that Germany applies the EU regulatory stack on top of national labor and data protection law. Its guide says works councils have codetermination rights under Betriebsverfassungsgesetz section 87 for the introduction and use of technical devices designed to monitor employee behavior or performance. It also says German employers must consult and reach agreement with the works council before deploying automated recruitment systems where one exists. That changes the rollout plan. A vendor pilot that starts as a quiet HR experiment can become a labor law problem if it monitors behavior or performance without the required consultation. JobCannon also lists GDPR Article 22 on automated decision making among the relevant recruitment rules, which means privacy counsel should not be invited after the workflow is already wired into the applicant tracking system. ## The practical preparation starts with inventory EDGE Certified Foundation's account of talent management makes the first step fairly prosaic: identify where AI touches recruitment, promotion, performance management, analytics, monitoring, or rewards. Then map which systems are merely administrative and which influence decisions about people. The latter group deserves the high risk file first, because those are the systems most likely to need documentation, oversight, and registration work. JobCannon's list of obligations gives builders and employers a clean starting checklist, even if nobody will call it fun. For each hiring AI system, assign an owner for the conformity assessment, collect technical documentation, define human oversight, and decide who is responsible for EU database registration. If a works council exists, build consultation into the deployment schedule rather than treating it as a post launch courtesy. The next useful signal to watch is not another webinar title. It is whether vendors serving German employers can hand over documentation that procurement, HR, privacy, and labor counsel can actually use. Since 2 August 2026, the polite answer to an AI recruiting pitch is simple: show the file. ## Sources - The EU Artificial Intelligence Act and Talent Management

Sources