Reducing the risk of missing or inadequate risk assessment means building a product where a compliant, standard‑aligned risk assessment (ISO‑12100 for machinery, EN 62368-1 for electronics, EN 71 for toys) is always produced, always complete, and always traceable to EU essential requirements. This is one of the most common CE‑compliance failures, so the mitigation strategy must be systematic, not ad‑hoc.
EU law makes risk assessment a mandatory part of CE conformity.
The Machinery Directive requires manufacturers to carry out a risk assessment and states machinery “shall not be considered to conform” without it (Annex I).
The Blue Guide emphasises that manufacturers must analyse risks and design accordingly.
Safety Gate recalls frequently cite “risk assessment missing” or “risks not sufficiently assessed” as grounds for withdrawal.
This makes risk assessment a high‑impact compliance risk—but also one of the easiest to control with the right workflow.
You’ll be reviewing product‑specific CE risk assessments generated with AI, so your role is to verify whether each assessment is complete, correct, and compliant for that exact product category. The review must focus on content accuracy, regulatory alignment, and technical feasibility.
Whether the AI correctly understands the product’s intended use, foreseeable misuse, operating environment, interfaces, and user groups.
Whether assumptions match the real product configuration and variants.
Whether all hazard types relevant to the specific product category are included (e.g., pinch points for machinery, EMC for electronics, ingestion hazards for toys).
Whether hazards are covered across the full lifecycle: transport, installation, operation, maintenance, cleaning, disposal.
Whether severity and probability ratings follow a defined method and are applied consistently.
Whether hazardous situations are described clearly and realistically.
Whether risk acceptability criteria are defined and used correctly.
Whether the legally required hierarchy is followed: inherently safe design → protective measures → information for use.
Whether proposed measures are technically feasible for the specific product.
Whether residual risks are identified, justified, and communicated.
Whether the correct directives/regulations are applied for the product category (e.g., Machinery Regulation, LVD, EMC, GPSR, RED, Toy Safety Directive).
Whether relevant harmonised standards are correctly referenced and interpreted (e.g., EN 60204‑1 for machinery, EN 62368‑1 for electronics, EN 71 for toys).
Whether each hazard and mitigation maps to the correct Essential Health and Safety Requirements (EHSRs).
Whether the risk assessment aligns with the actual design, drawings, components, safety functions, testing plan, instructions, and labelling.
Whether no contradictions exist (a common LLM failure mode).
Whether every hazard has a corresponding mitigation and residual risk.
Whether the document is structured, traceable, and auditable.
Whether design decisions and assumptions are transparent and justified.
Whether the AI invented standards, directives, test methods, or technical features.
Whether the AI misunderstood the product’s technology or operating principles.
Whether the AI omitted foreseeable misuse or lifecycle hazards typical for that product category.
Essential Health and Safety Requirements (EHSRs) of the applicable directives/regulations.
Obligations in the Blue Guide (risk analysis, documentation, conformity).
Product‑specific legal frameworks (e.g., Machinery Regulation, GPSR, RED, LVD, EMC).
ISO 12100 for methodology.
Type‑B and Type‑C standards relevant to the product category.
Standards governing electrical safety, control systems, EMC, mechanical strength, chemical safety, etc.
Completeness, accuracy, lifecycle coverage, and traceability.
Consistency with the technical file and Declaration of Conformity.
Evidence‑based justification of risk‑reduction measures.