One in four large Spanish companies expects to cut jobs because of AI

A regulation on artificial intelligence at work would follow the general legislative procedure: there is no distinct parliamentary process because it involves AI. The key is to distinguish between approving a law, amending existing labor legislation, and developing a law through regulatory provisions, which are not approved through the same parliamentary route.

1. Presentation of the initiative

The regulation can begin through different channels:

• Government bill: the Executive submits a text to the Congress accompanied by a statement of reasons and background.
• Bill proposal: it can come from a parliamentary group or a deputy with fourteen other signatures. Initiatives can also be promoted by the Senate, regional assemblies, and citizens, meeting applicable requirements.

Proposals generally require consideration: the Plenary decides whether to start processing them. Those coming from the Senate that have already passed this stage do not repeat it in the Congress. The Government may express its opinion and, when appropriate, its budgetary disagreement.

2. Admission and amendments in the Congress

The Board qualifies the initiative, orders its publication, and refers it to the competent committee according to its content. A labor regulation on AI could combine labor and technological issues: the Board may assign the matter to one committee and request reports from others.

For bills, the initial amendment period is fifteen working days, subject to extension or reduction. Changes, deletions, or additions to the articles may be proposed. Groups may also submit total amendments to return the bill to the Government or replace it with another text. If its return is approved, the bill is rejected.

3. Working group, committee, and approval

A working group, formed within the committee, studies the text and amendments and drafts a report. Then, the committee debates article by article and approves an opinion, that is, the text resulting from its work.

The committee may request documentation and appearances from experts, unions, business organizations, or authorities for advice. This participation informs the debate but does not replace the parliamentary decision.

The Plenary debates the opinion and the maintained amendments. However, in delegable matters, the committee may approve the law with full legislative competence, without a final Plenary vote, unless the Plenary recovers that decision.

The rule for an ordinary law is a simple majority: more favorable than opposing votes, with the required quorum. If the content belongs to matters reserved for organic law, an absolute majority of the Congress is required in a final vote on the whole; regulating labor AI does not automatically make a law organic.

4. Senate review

The Senate receives the text approved by the Congress and may approve it without changes, introduce amendments, or oppose a veto. It generally has two months; in urgent procedure, the period is reduced to twenty calendar days.

Modifications return to the Congress, which decides whether to incorporate them by simple majority. To lift a veto, the Congress needs an absolute majority or, after two months since its imposition, a simple majority. Thus, the Senate participates in the review, but the Congress can resolve the discrepancy.

5. Urgent route and European alignment

Another possibility is a royal decree-law, subject to the constitutional requirement of extraordinary and urgent need and its material limits. The Congress must decide its validation or repeal within thirty days. If validated, it may agree to process it as an urgent bill, opening the possibility of amendments and subsequent Senate intervention.

The national content must respect applicable European law. European regulations are directly applicable; directives require national incorporation. This coordination conditions what Spain can regulate but does not replace the parliamentary procedure. The described phases are included in the Congress Rules.

The Bank of Spain does have powers to prepare reports on the labor impact of new technologies, including artificial intelligence, when it analyzes it as an economic phenomenon. Its role is to research, produce statistics, and advise: not to regulate labor relations or decide the State's technological policy. The authorization derives from its general powers, not from an exclusive competence over technology and employment.

Legal basis

The main basis is article 7.5.e) of Law 13/1994, on the Autonomy of the Bank of Spain, which expressly assigns it the function of "advising the Government, as well as preparing reports and studies as appropriate." This provision allows studying economic transformations affecting employment, wages, or productivity, although the law does not literally mention automation or artificial intelligence. It can be consulted in the Autonomy Law.

Article 7.5.f) adds the power to "prepare and publish statistics related to its functions" and assist the European Central Bank in collecting statistical information. Meanwhile, article 7.2 establishes that it will support the Government's general economic policy, without prejudice to price stability and its functions within the European System of Central Banks.

What it can analyze and publish

These powers support reports examining how the incorporation of new technologies modifies the functioning of the economy and the labor market. Possible contents include:

• The effects of automation on employment, task substitution, and creation of new activities.
• Changes in wages, productivity, demand for qualifications, and professional adaptation needs.
• Differences between sectors, companies, and worker groups.
• Impacts on growth, inflation, competitiveness, and corporate financial situation.

It is not only a legal possibility. The presentation of the 2023 Annual Report, on April 23, 2024, addressed technological changes and the adaptation of professional qualifications. Additionally, research priorities announced in December 2025 include the labor market and technological transition, including artificial intelligence, as reported by Demócrata.

Collaboration and use of data

The Bank can rely on agreements to obtain information and develop research linked to its functions. An example is the agreement with Red.es, published in August 2026, to evaluate and analyze digitalization and advise on the transformation of SMEs and self-employed workers.

It also announced on February 9, 2026, an agreement with the OEPM to facilitate economic studies on innovation through patent and utility model data. Access is carried out in a controlled environment that protects confidentiality. These agreements facilitate research but do not themselves expand the Bank's legal powers.

Limits and value of its conclusions

The competence is analytical and consultative, not executive in labor matters. Therefore, a report may formulate diagnoses, scenarios, and recommendations, but does not allow the Bank to set wages, modify working conditions, impose professional reconversion measures, or authorize or prohibit technologies due to their labor effects.

Studies must maintain connection with its economic functions and respect confidentiality, professional secrecy, and data protection rules. Its recommendations are not legally binding: they provide evidence for decisions by competent authorities. In short, the Bank can study how technology transforms employment; that faculty does not make it a labor authority or general regulator of artificial intelligence.

Companies incorporating artificial intelligence into their production processes in Spain must comply with the European AI Regulation, data protection, and applicable labor, industrial safety, and sectoral regulations. The requirements depend on the specific use, its risks, and whether the company develops the system or simply uses a third-party solution: not all industrial AI is subject to the same obligations.

1. Identify the use and responsibilities

The first step is to determine the system's purpose, the data used, and the people affected. It must also be distinguished whether acting as a provider, manufacturer, importer, or professional user. A productivity optimization tool does not automatically receive the same treatment as one intended to select workers or control safety functions of a machine.

For high-risk systems, the European framework contemplates risk management, data quality, technical documentation, traceability, human oversight, accuracy, robustness, and cybersecurity. The conformity assessment, CE marking, and European registration apply when appropriate, not to every company using AI.

2. Separate current and future obligations

From February 2, 2025, prohibitions on certain uses and the obligation to ensure sufficient AI knowledge for personnel handling it apply. Among relevant prohibitions is inferring emotions at work, except for medical or safety exceptions. Training must be adapted to functions and context, as explained by AESIA on literacy.

Transparency obligations apply from August 2, 2026. According to the timeline explained by Demócrata, the European reform postponed full high-risk obligations to December 2, 2027, for Annex III uses and to August 2, 2028, for systems incorporated into regulated products. These postponements do not suspend other applicable rules.

3. Protect personal data

If data of workers, clients, or other persons are processed, the GDPR and Organic Law 3/2018 must be respected. This requires:

• Having a legal basis and explaining the purposes of processing; consent is not always the only possible basis.
• Limiting data and its retention, ensuring its security, and allowing the exercise of rights.
• Contractually regulating provider access and international transfers when they exist.
• Conducting an impact assessment when processing may entail high risk.

Decisions that are exclusively automated with legal effects or similarly significant consequences must respect the conditions and guarantees of article 22 of the GDPR, including human intervention when appropriate.

4. Respect labor rights and safety

Implementation must respect information and consultation obligations to workers' representatives when affecting employment, organization, control, or evaluation. Collective agreements must also be reviewed and discriminatory uses avoided.

In industrial environments, the Occupational Risk Prevention Law, the Industry Law, and machinery regulations continue to apply. Incorporating AI does not replace equipment safety requirements or preventive measures.

5. Transparency, contracts, and controls

When appropriate, interaction with AI must be notified and artificial or manipulated content indicated, according to transparency rules. Likewise, licenses, copyrights, and confidentiality must be respected, and sectoral cybersecurity obligations checked.

As practical preparation, it is advisable to inventory systems, assign responsibilities, review providers, and keep evidence. The AESIA practical guides help organize compliance, without replacing legal obligations.