THE BOTTOM LINE
- Regulatory Preview: Spain’s cautious, human-centric approach signals the likely regulatory mood across the EU for high-stakes AI. Legal tech and AI developers must prioritize transparency and human oversight in their products to gain traction in the legal sector.
- Risk Mitigation in Focus: The explicit ban on using AI for profiling or risk assessment within the judiciary reinforces the ‘high-risk’ designation under the EU AI Act, impacting how companies can use similar tools in regulated sectors like finance and HR.
- Litigation Remains Human: While AI may streamline judicial back-office tasks, final legal judgments remain firmly in human hands. Businesses should not expect AI to replace the need for compelling legal arguments aimed at a human decision-maker.
THE DETAILS
Spain’s General Council of the Judiciary (CGPJ), the governing body of the country’s judges, has issued a landmark instruction on the use of artificial intelligence in judicial activities. The directive’s core principle is unambiguous: effective human control. The guidelines establish that AI systems, particularly generative AI, are to be used only as support tools. They cannot operate autonomously to make judicial decisions, assess facts or evidence, or interpret the law. This ensures that judges and magistrates retain full, conscious control and remain exclusively responsible for their rulings, safeguarding judicial independence and accountability in an age of automation.
The instruction draws a clear line between permitted and prohibited uses. Judges are allowed to use AI for tasks like legal research, analyzing and structuring case information, or creating internal summaries and work drafts. While they can use AI to generate drafts of judicial decisions, these are strictly considered support instruments. Any such draft requires a “complete and critical personal review and validation” by the judge. Conversely, the rules strictly forbid using AI to automate or delegate decision-making, profile individuals, predict behavior, or assess risk. Furthermore, any content generated by AI cannot be incorporated into a final ruling without the judge’s thorough personal evaluation and sign-off.
Crucially, the CGPJ has created a “walled garden” for AI use. Judges are only permitted to use AI applications that are provided by competent justice administrations and have been vetted by the CGPJ itself. This measure is designed to prevent the use of public, unverified AI tools that could compromise the confidentiality of sensitive legal data, introduce algorithmic bias, or produce unreliable outputs. This approach aligns with the broader principles of the new EU AI Act and demonstrates a proactive step by a major European judiciary to harness the efficiency of AI while building a framework of trust, security, and ethical use.
SOURCE
Source: Spain’s General Council of the Judiciary (CGPJ)
