Ethics and the Future of Legal AI
EU AI Act, CGPJ Instruction 2/2026, ethical limits, agentic AI — the debate that matters.
The ethical debate: Can AI practice law?
The short answer is no. The long answer explains why, and what implications it has for your professional practice.
What AI can do
- Process information faster than any human
- Find patterns in thousands of rulings
- Generate draft legal documents
- Work 24/7 without rest
What AI CANNOT do
- Exercise moral judgment: has no values, beliefs, or sense of justice
- Understand human context: doesn't know your client just lost their father and needs empathy
- Assume responsibility: if something goes wrong, AI doesn't lose its license
- Make ethical decisions: the "recommend settlement vs go to trial" dilemma requires professional judgment
The EU AI Act
The Regulation (EU) 2024/1689, known as the AI Act, is the world's first comprehensive AI regulation. In force from August 1, 2024.
Risk classification
| Level | Description | Legal example |
|---|
| Unacceptable | Manipulative or discriminatory AI | Social scoring, mass surveillance |
| High risk | AI affecting fundamental rights | Predictive justice, credit scoring |
| Limited risk | AI with transparency obligations | Legal chatbots, document assistants |
| Minimal risk | No special restrictions | Spell checkers, email classification |
Implications for legal practice
- Predictive justice systems: if AI predicts case outcomes, may be "high risk" classified.
- Document assistants: legal chatbots must disclose AI nature to users.
- Automated contract analysis: depending on consequences, could be limited or high risk.
Algorithmic bias in law
What it is
Algorithmic bias occurs when AI produces systematically unequal results for certain groups. In the legal context, this is especially serious.
How to protect yourself
- Question results: if AI suggests something that "sounds unfair", it may be reproducing a bias.
- Diversify sources: don't depend solely on AI for your legal reasoning.
- Know the data: ask what data the tool you use was trained on.
- Think critically: would the AI's recommendation benefit your client if they belonged to another demographic group?
The future of the legal profession
Realistic 5-year scenario (2026-2031)
| Area | Expected evolution |
|---|
| Research | AI as mandatory first step; manual search only for very specific topics |
| Drafting | AI generates 80% of first draft; lawyer perfects |
| Negotiation | AI analyzes positions and suggests strategies; negotiation remains human |
| Litigation | AI prepares all documentation; oral hearings remain in-person |
| Client service | Chatbots handle 60% of initial queries; lawyer for complex matters |
What WILL happen
- Firms that don't adopt AI will lose competitiveness.
- New specialties: lawyers specialized in AI, tech regulation, algorithmic auditing.
- Billing changes: fewer hours billed per case, but more value per case.
- Democratization of access to justice.
Ethical code for AI use in your firm
The 10 principles
- Transparency: always inform the client when using AI
- Verification: never present AI results without verification
- Responsibility: the signature is mine, the responsibility is mine
- Confidentiality: client data never goes to services without DPA
- Training: continuously train on AI and its limitations
- Supervision: AI doesn't make decisions; I make them with AI support
- Equity: question results that may be biased
- Proportionality: use AI where it adds value, not for technology's sake
- Documentation: record what AI tools I use and for what
- Updates: periodically review if my practices remain adequate
Module summary
| Topic | Key point |
|---|
| AI and ethics | AI has no moral judgment; the lawyer always decides |
| AI Act | First comprehensive regulation; affects legal tools |
| Bias | AI can reproduce historical biases; question results |
| Future | AI doesn't replace lawyers; replaces routine tasks |
| Ethical code | 10 practical principles for your firm |