EU AI Act: what students and grads should know now
The EU AI Act's biggest deadline just moved 16 months

The EU AI Act's biggest deadline just moved by 16 months.
and yet a whole new set of rules switched on last week.
Both things are true at once. The rules for high-risk AI slipped from August 2026 to December 2027, whilst the transparency rules started on 2 August. If you're a student or new grad, the EU AI Act already touches two systems you'll definitely meet: the ones that mark your work and the ones that screen your CV.
The EU AI Act timeline so far
The Act entered into force on 1 August 2024, then switched on in stages (Article 113):
- 2 February 2025: banned practices and the AI literacy duty
- 2 August 2025: rules for general-purpose AI models (the big models behind chatbots)
- 2 August 2026: most of the rest, including transparency
- 2 December 2027: high-risk systems in sensitive areas, like education and hiring (delayed)
- 2 August 2028: high-risk AI built into regulated products (delayed)
What the Digital Omnibus changed
Those last two dates are new.
The EU's "Digital Omnibus on AI" (a package of amendments meant to simplify the rules) was published in the Official Journal on 24 July as Regulation (EU) 2026/1744, and came into force three days later (Cuatrecasas). It:
→ pushed stand-alone high-risk rules from 2 August 2026 to 2 December 2027 → pushed high-risk AI inside products to 2 August 2028 → gave generative AI systems already on the market until 2 December 2026 to start marking AI-made content → added bans on non-consensual sexual deepfakes and AI-generated child abuse material → softened AI literacy: firms must take measures to support staff literacy, but don't have to guarantee any level for any individual (Addleshaw Goddard)
What started on 2 August 2026
The transparency rules (Article 50) now apply (Cooley):
- tell people when they're talking to an AI
- mark synthetic content so it can be detected
- give notice when emotion recognition or biometric categorisation is used
- disclose deepfakes and AI-written text published on matters of public interest
Fines go up to €15 million or 3% of worldwide turnover, whichever is higher. The Commission adopted guidelines on Article 50 on 20 July, so firms aren't flying completely blind.
Why the EU AI Act matters for students and grads
Here's the bit nobody puts in the freshers' guide.
Annex III lists the high-risk areas, and two of them are basically your next five years:
→ education: AI for admissions, evaluating learning outcomes, deciding your level of education, and monitoring students during exams → employment: AI for recruitment and selection, including tools that screen applications and rate candidates
And one thing is already banned: AI that infers emotions in workplaces and education institutions, outside medical or safety reasons (Article 5).
So the delay matters. Exam proctoring and CV screening tools now get until December 2027 before the full high-risk rules apply. We covered how screening already works in what employers actually test for, and how governments are deploying AI in public services.
What uni won't teach you about AI regulation
→ law modules treat the AI Act as a reading list, not a calendar. The dates are half the story → business and computer science degrees rarely mention it, even though graduates will build and buy these systems → "regulation" gets taught as a brake, when in practice it's a set of job specs: someone has to do the documentation, testing and disclosure
The critical-thinking move: 4 questions for any AI system you meet
- Which bucket is it in? Banned, high-risk, transparency-only, or minimal? Grading and hiring tools are high-risk
- Who's the provider and who's the deployer? The company that builds it and the one using it on you have different duties
- What must I be told? Since 2 August you should know when you're talking to an AI or looking at a deepfake
- When does the rule actually bite? A rule on paper and a rule in force are different things. Check the date
What to do this month
- Learn the five dates above. Honestly they'd beat most answers in an interview at a policy, legal or tech firm
- Check your uni's tools. Find out whether your university uses AI for proctoring or marking, and what it tells students about it. Our piece on how students actually use AI at uni is a good starting point
- Ask employers about screening. If you're applying to EU employers, ask how AI is used in their hiring process
- Read one primary source. Article 50 is short. Read it in full, then compare it with a firm's summary
More in our public sector and policy hub.
so TLDR: the high-risk rules slipped to December 2027, but transparency is live now, and the systems marking your work and screening your CV are firmly in scope.
Would you trust an AI to mark your exam if the rules protecting you didn't apply until 2027?
Sources
- Digital Omnibus on AI has been publishedCuatrecasas · 24 JUL 2026
- EU AI Act: AI Omnibus formally adoptedAddleshaw Goddard · 01 JUL 2026
- EU AI Act: Transparency Obligations Take Effect 2 August 2026Cooley · 03 AUG 2026
- Article 113: Entry into Force and ApplicationEU Artificial Intelligence Act (artificialintelligenceact.eu) · 12 JUL 2024
- Annex III: High-Risk AI Systems Referred to in Article 6(2)EU Artificial Intelligence Act (artificialintelligenceact.eu) · 12 JUL 2024
- Article 5: Prohibited AI PracticesEU Artificial Intelligence Act (artificialintelligenceact.eu) · 12 JUL 2024
Get the next issue by email
theVANTAGEd is written by students to bring critical thinking to AI across real industries. One issue every two weeks.