AI in law: what it can do, what it cannot do and what risks it carries

Artificial intelligence is already in Chilean legal practice: it drafts contracts, analyses case law, identifies problematic clauses in seconds.…

2026-06-10 · Automation

Artificial intelligence is already in Chilean legal practice: it drafts contracts, analyses case law, identifies problematic clauses in seconds. The tools exist and some firms are using them. But using them well —and in a legally safe way— requires understanding what they can do, what they cannot do, and which responsibilities do not disappear just because an algorithm is involved.

What AI does in the legal world today

The most frequent applications of AI in legal practice are:

Contract review and analysis: language models trained on legal texts can identify unusual clauses, compare terms with market standards or flag potential risks in contracts running to many pages, in a few minutes.
Legal research: semantic search across databases of case law, doctrine and legislation. Instead of searching for exact keywords, the system understands the legal question and returns relevant sources.
Assisted drafting: generating drafts of contracts, claims, pleadings or standard clauses from parameters defined by the lawyer.
Due diligence: mass review of documents in acquisition or merger processes, identifying relevant milestones, contingent obligations or hidden risks.
Outcome prediction: models that analyse case law to estimate the probability of success in a case, albeit with important limitations.

What AI cannot do (yet)

It pays to be realistic. Current AI systems, however sophisticated, have important limitations in a legal context:

They do not take on professional responsibility. The lawyer who signs a pleading is responsible for its content, even if a language model generated it. AI is a tool, not a qualified professional.
They can "hallucinate". Language models sometimes generate citations to case law that does not exist or interpret rules incorrectly. Every output must be critically reviewed by a professional.
They do not know the negotiating context. A contract is not just text: it has history, power relationships, unwritten intentions. AI reads the document, not the relationship.
Chilean legislation is thin in their training data. Most legal AI models were trained mainly on English-language data or common law. Applying them to Chilean law requires specific validation.

The legal risks of using AI without judgement

Delegating to AI without proper supervision can create concrete problems:

Confidentiality: uploading client documents to external AI platforms can breach professional privilege and, from December 2026, the duties of Ley 21.719 if the provider processes that data without adequate safeguards.
Undetected errors: a clause generated incorrectly that passes without review can lead to litigation or significant financial loss.
Bias in research: if the model selects case law in a biased way, it can lead to poorly founded legal strategies.

A concrete example: reviewing a supplier contract

A medium-sized company receives a 60-page contract from a foreign technology supplier. Previously, reviewing it took days of a lawyer's work. With a legal AI tool, in 10 minutes it can obtain a summary of the most relevant clauses, the risk points (liability limitations, governing jurisdiction, penalties) and a comparison with market standards. The lawyer then reviews those identified points and negotiates on that basis. The review time falls drastically; the lawyer's responsibility does not.

Which regulatory framework applies

The use of AI in law in Chile intersects with at least two regulatory frameworks that already exist or are in progress:

Ley 21.719 (personal data): if the AI processes data of clients or parties to proceedings, data protection obligations apply. Fully in force: December 2026.
AI bill: if the system assists in decisions that significantly affect people —such as a legal risk assessment determining access to credit or litigation strategy— it could fall into the high-risk category under the bill before the Senate.

In summary

AI in law is a reality, not a future promise. Its efficiency benefits are real and available today. But professional responsibility is not assumed by the algorithm: it is assumed by whoever uses it and signs the result. Using it well means knowing what it can do, verifying what it produces and keeping legal judgement where AI does not reach.

General information on technology and legal practice. Informational content: it does not replace professional advice.

— Mindset & Code · Legal-Tech Chile · law and technology, explained simply.