Executive summary
The European Commission announced that the AI Omnibus entered into force on 27 July 2026. The reform is presented as a targeted simplification of the EU AI rulebook that preserves safeguards while easing compliance, extending some timelines and supporting innovation.
One of the most important changes for democratic resilience is the treatment of AI literacy. According to the Commission’s summary, the previous company-level obligation is replaced by a non-binding encouragement, while the Commission and Member States assume a stronger role in promoting AI literacy.
This changes the accountability structure. A duty imposed on providers and deployers can be assessed through organisational records and enforcement. A general public-policy commitment to promote literacy requires different evidence: programmes, target groups, budgets, participation, learning outcomes and inclusion.
Core forensic question
When a legal duty becomes an encouragement, who remains accountable for ensuring that people can use and evaluate AI systems safely and critically?
The issue is not whether simplification is inherently harmful. It is whether the replacement mechanism can produce measurable and equitable literacy outcomes.
The accountability shift
A direct organisational duty creates an expected evidence trace:
- internal training policies;
- staff competence frameworks;
- records of completed training;
- risk-based learning requirements;
- role-specific guidance;
- audit and enforcement records.
A public obligation to promote AI literacy creates a different trace:
- national strategies;
- public budgets;
- curricula;
- funded programmes;
- accessibility standards;
- participation data;
- learning assessments;
- reporting by demographic and occupational group.
If neither trace is clearly required, AI literacy may become a policy aspiration without an accountable delivery mechanism.
Why AI literacy matters
AI literacy is not limited to learning how to operate a chatbot. In a democratic and administrative context, it includes the ability to:
- understand that AI output may be inaccurate;
- distinguish generated content from evidence;
- identify when an automated system affects a decision;
- recognise bias, uncertainty and manipulation;
- protect personal and confidential information;
- request human review;
- challenge an AI-assisted administrative outcome;
- understand the limits of automated legal or factual analysis.
A framework focused only on technical adoption would not be sufficient.
Quantitative-formalism risks
Governments and institutions may report:
- number of training sessions;
- number of participants;
- number of online modules;
- number of organisations reached;
- number of guidance documents.
These outputs do not establish that participants can critically assess AI systems.
A more meaningful measurement framework should include:
- completion rates;
- pre- and post-training assessment;
- ability to identify unreliable output;
- ability to recognise when human review is required;
- accessibility across languages and disability groups;
- coverage of vulnerable or digitally excluded populations;
- retention of knowledge over time;
- use of complaint and remedy mechanisms.
Alternative explanations
A flexible, non-binding approach may allow training to be adapted to different sectors and technologies. It may also reduce disproportionate burdens on smaller organisations.
Existing obligations under labour law, professional standards, data protection, consumer protection or sectoral regulation may still require competence and training in particular contexts.
The effect of the reform therefore depends on the full legal text, implementing measures and national programmes. A change in wording does not by itself prove that literacy outcomes will decline.
Public-interest test
To assess the reform, the Commission and Member States should publish answers to five questions:
- Which population groups are expected to receive AI-literacy support?
- Which institutions are responsible for delivery?
- What minimum competencies are expected?
- How will outcomes be measured?
- What happens where insufficient literacy creates a risk to rights or public services?
Preliminary finding
The AI Omnibus appears to shift AI literacy from a directly attributable organisational obligation toward a broader policy-promotion model.
That may increase flexibility, but it also diffuses responsibility. The success of the change should not be measured by the number of campaigns or training offers. It should be measured by demonstrable competence, equitable access and the ability of people to recognise and challenge harmful AI use.
Recommended Civic Forensics modules
- Analysis of Regulatory Change
- Quantitative Claims Module
- Detector of Quantitative Formalism
- Expected Evidence Trace Plan
- Analysis of AI-Assisted Public Administration
- Legal and Ethical Publication Review
Source
European Commission, AI Omnibus enters into force , 27 July 2026: https://digital-strategy.ec.europa.eu/en/news/ai-omnibus-enters-force
Limitations
This analysis relies on the Commission’s public summary of the reform. A complete legal assessment requires the final legislative text, amended provisions, recitals, transitional rules and relevant implementing measures.