Short answer: high-risk AI systems are those the EU AI Act considers able to seriously affect people’s health, safety or fundamental rights, such as AI that screens job applicants, scores creditworthiness or decides access to education. They face the strictest rules in the Act. Most websites don’t run high-risk AI. After the 2026 Digital Omnibus, the high-risk obligations apply from 2 December 2027 for stand-alone systems and 2 August 2028 for AI built into regulated products.
Two routes to “high-risk”
1. AI in regulated products (Annex I)
AI that is a safety component of, or itself is, a product covered by EU product safety laws that require third-party conformity assessment, such as medical devices, machinery, toys, lifts, vehicles and radio equipment. These obligations apply from 2 August 2028.
2. Stand-alone uses listed in Annex III
AI used in these areas, from 2 December 2027:
- Biometrics: remote identification, biometric categorisation by sensitive traits, emotion recognition (where not banned).
- Critical infrastructure: safety components for traffic, water, gas, heating and electricity.
- Education: admission, assessing learning outcomes, proctoring exams.
- Employment: recruiting, screening and ranking applicants, decisions on promotion, termination and task allocation, monitoring performance.
- Essential services: creditworthiness of individuals, life and health insurance pricing, eligibility for public benefits, emergency call triage.
- Law enforcement, migration and border control, justice and democratic processes.
The exception
An Annex III system is not high-risk if it doesn’t pose a significant risk of harm, for example because it only performs a narrow procedural task, improves the result of a human activity already completed, or does preparatory work for a human assessment. That exception never applies when the system profiles individuals. A provider relying on it must document the assessment.
What high-risk providers must do
- Risk management system across the lifecycle.
- Data governance for training, validation and testing data.
- Technical documentation and automatic logging.
- Transparency and instructions for use for deployers.
- Human oversight measures built in.
- Accuracy, robustness and cybersecurity.
- Quality management system, conformity assessment, CE marking and registration in the EU database.
What deployers of high-risk AI must do
- Use it according to the provider’s instructions.
- Assign trained people to oversee it.
- Monitor it and keep the logs it generates.
- Inform workers and their representatives before using it at work.
- Inform people that they’re subject to a high-risk AI decision, where required.
- In some cases, carry out a fundamental rights impact assessment.
Does your website use high-risk AI?
| Website feature | High-risk? |
|---|---|
| Support chatbot, AI search, product recommendations | No. Chatbots have transparency duties under Article 50. |
| AI-written blog posts and AI images | No. Article 50 labelling rules may apply. |
| Careers page with AI that ranks or rejects applicants | Yes, employment is in Annex III. |
| Loan or buy-now-pay-later form with your own AI credit scoring | Likely yes, creditworthiness of individuals is in Annex III (fraud detection is excluded). |
| Online course platform with AI grading or admission decisions | Likely yes, education is in Annex III. |
| Insurance quote tool with AI pricing for life or health cover | Likely yes. |
| Webcam feature inferring visitors’ emotions | Possibly high-risk, and banned in workplaces and schools. |
Two practical notes. Buying a recruiting or credit tool from a vendor makes the vendor the provider and you the deployer, with lighter but real duties. And if you put your own name on a high-risk system or substantially modify it, you can become its provider.
What the Digital Omnibus changed
The Digital Omnibus on AI (Regulation (EU) 2026/1744) moved the high-risk deadlines from 2 August 2026 to 2 December 2027 (Annex III) and 2 August 2028 (Annex I), because harmonised standards and national authorities weren’t ready. It’s a delay, not a cancellation. It did not delay the bans, which apply since February 2025, or the Article 50 transparency duties, which apply since 2 August 2026.
Penalties
Breaches of high-risk obligations fall in the 15 million euros or 3% tier, with the lower cap for SMEs. See EU AI Act fines and penalties explained.
What to do now
- Run a website AI audit and flag anything touching hiring, credit, insurance, education or biometrics.
- For flagged tools, ask the vendor about its high-risk compliance plan for December 2027.
- For everything else, focus on what applies now: the Article 50 transparency rules.
For the transparency side on WordPress, the free EU AI Act Ready plugin handles chatbot notices, AI content and image labels, and keeps a record of your decisions. It doesn’t make a high-risk system compliant; that needs the provider’s conformity process.
Frequently asked questions
Is ChatGPT a high-risk AI system?
No. It’s built on a general-purpose AI model, which has its own rules. It becomes part of a high-risk use only if someone deploys it for a high-risk purpose, such as screening job applicants.
Is a chatbot high-risk?
Not normally. Chatbots have transparency duties. A chatbot that decides who gets a loan or a job could be part of a high-risk system.
When do high-risk rules apply?
2 December 2027 for Annex III systems and 2 August 2028 for AI in regulated products. See the full AI Act timeline.