Spain opens AI360 process to frame a social contract on artificial intelligence

Spanish Prime Minister Pedro Sánchez launched the AI360 process on September 21, setting out a twelve-month exercise intended to shape how Spain manages artificial intelligence across work, public oversight, infrastructure and rights. The announcement does not itself create a new law. It starts a policy process in which the government says it will convene social partners next month to work on a new social contract, with job security as the central theme.

The change is political and procedural rather than legislative. Spain is moving from broad concern about artificial intelligence to a structured national discussion with a fixed duration and a list of priority issues. Sánchez linked the process to employment, legal responsibility, deepfakes, young people’s mental health, intellectual property, data-centre infrastructure and public supervision. That framing matters because it treats AI not only as a technology agenda, but also as a labour, liability and infrastructure question.

A process before a law

The AI360 initiative is not the same as Spain’s national AI bill. The bill is not expected before the end of 2027, which means the current process sits in the gap between political agenda-setting and formal legislation. Its practical role is therefore to gather positions, test policy priorities and define the terms of a future settlement before the government reaches the stage of a national legal text.

The twelve-month duration gives the process a defined timetable, but it does not prove that consensus will be reached. It sets a window for consultation and policy design. The planned meeting with social partners next month indicates that labour-market effects are the first concrete area for negotiation. In this context, “social contract” means an attempt to align employers, workers and government around rules or commitments for the use of AI in the workplace, especially where job security is concerned.

Sánchez expects more than half of Spanish companies to use AI by 2030. That number is a government expectation, not an independent measurement of current adoption and not a benchmark produced by a technology vendor. It supports the government’s case that AI will become a mainstream business tool, but it does not show how companies will use AI, which sectors will adopt it fastest, or whether adoption will translate into productivity gains, job losses or new tasks.

Jobs, responsibility and harms

The employment focus is the clearest practical implication for companies and workers. If more firms introduce AI systems, questions will arise over task automation, worker monitoring, recruitment tools, training and the distribution of gains. The AI360 process signals that Spain wants job security discussed before a wider legal framework is in place. It also suggests that workplace AI may become a matter for negotiation with social partners, not only for internal corporate technology teams.

Sánchez also called for legal and financial responsibility for algorithmic harm. That moves the discussion beyond voluntary ethics statements. The issue is who carries responsibility when an AI system produces damage, whether through discriminatory outcomes, erroneous decisions, unsafe outputs or other harmful effects. The announcement does not specify the mechanism, threshold or enforcement model. It does, however, identify liability as a core component of the proposed social settlement rather than a secondary technical matter.

Deepfakes, young people’s mental health and intellectual-property violations form another part of the agenda. These are different risks, and the announcement groups them under the wider need for stronger oversight. Deepfakes raise questions of manipulation and authenticity. Mental-health concerns point to the effects of AI-mediated services or content on younger users. Intellectual-property violations concern the use or reproduction of protected work. The AI360 process does not resolve these questions, but it places them on the national policy timetable.

Infrastructure becomes part of AI policy

The plan also includes an AI gigafactory, models developed with the Barcelona Supercomputing Center, vouchers for small businesses and self-employed workers, curriculum updates and stronger oversight. These elements show that the Spanish government is not treating regulation and industrial capacity as separate tracks. The agenda combines public capability, business adoption, skills and supervision, although the announcement does not provide implementation details or results against which those measures can yet be assessed.

The proposed AI gigafactory and the models developed with the Barcelona Supercomputing Center point to infrastructure and national capability. The available facts do not indicate the scale, cost or technical specifications of the gigafactory, nor the intended uses of the models. What can be said is that the government sees computing capacity and model development as part of Spain’s AI strategy, rather than relying only on rules for private-sector deployment.

Sánchez said data centers should be assessed for water and electricity demands. This is a significant addition to the policy frame because it connects AI deployment with resource constraints. AI systems depend on data infrastructure, and data centers can require substantial water and electricity. The statement does not provide measurements of expected consumption or set limits. It does, however, indicate that infrastructure approvals or assessments could become part of the national debate over AI expansion.

What the numbers do and do not show

The numerical claims in the announcement are limited. The twelve-month process proves that the government has set a period for consultation and policy development. The planned convening of social partners next month shows the first operational step. The expectation that more than half of Spanish companies will use AI by 2030 is a forecast from the prime minister, not evidence of present market penetration. The absence of a national AI bill before the end of 2027 shows that binding national legislation remains some distance away.

There are no publisher benchmarks, product performance tests or independent measurements in the facts provided. The Barcelona Supercomputing Center reference concerns model development, not a reported benchmark result. The company-adoption figure is a political projection, not a vendor claim and not an independently verified adoption rate. For readers and policymakers, the distinction is important: the announcement supports analysis of Spain’s policy direction, but it does not allow conclusions about technical performance, economic impact or actual deployment levels.

Internationally, Spain favors strengthening the EU AI Office and the UN scientific panel rather than creating another institution. That position places the AI360 process within existing European and global governance structures. The practical implication is that Spain is not presenting its national initiative as a substitute for broader coordination. Instead, it is linking domestic consultation with support for institutions already tasked with AI oversight or scientific assessment. The unresolved question is how the year-long process will translate into enforceable rules once the national bill eventually arrives.