The responsibility gap: who governs AI in legal practice in LATAM?

AI adoption in Latin American law firms is accelerating faster than most firms’ governance can keep up with. Regional legal tech spend is on a steep growth curve, and lawyers across the region are already using AI daily, in some markets for the majority of firms with more than a handful of lawyers. Governance frameworks, internal policies, and clear lines of accountability have not kept pace with that curve. That gap — between how fast firms are adopting AI and how clearly they can answer “who is responsible when it goes wrong” - is the one that matters most right now.

Regulation is arriving, unevenly

Unlike the EU, Latin America is not moving toward a single, horizontal AI law. Most observers expect a more gradual, sector-by-sector approach, shaped country by country rather than by one regional standard. In that vacuum, it’s the bar associations — not legislators — who have moved first.

Mexico’s national bar association, the Barra Mexicana, published the region’s first formal guidelines for the responsible use of AI in legal practice at the end of 2025, developed with input from a legislative AI commission. Chile’s Colegio de Abogados followed with its own guidance on professional AI use, built around existing duties under its code of ethics. Argentina’s Colegio Público de Abogados de la Capital Federal has issued similar principles, anchored in transparency toward clients about when AI tools are used.

Different institutions, same throughline: confidentiality, transparency with clients, and — above all — a firm line that human supervision cannot be delegated to a tool.

‍

The duty never moves to the machine

This is the part firms sometimes miss when they focus on the technology and not the practice around it: no guideline anywhere in the region - or globally - shifts professional responsibility onto the AI. The lawyer remains accountable for every citation, every draft, every piece of advice, regardless of what generated the first version. Courts elsewhere have already shown what happens when that duty is treated as optional: sanctions, disqualifications, and referrals to disciplinary bodies, in cases where AI-generated content was submitted without verification.

For firms in LATAM, that means the real question isn’t “should we adopt AI” — that decision is largely already made by client expectations and competitive pressure. The real question is: can we show, at any moment, who is accountable for what the system produced, where the underlying source came from, and how it was verified?

‍

Where the gap actually shows up

In practice, the responsibility gap tends to appear in three places:

• Traceability. If a piece of legal research can’t be traced back to an official source, verifying it — and defending it — becomes guesswork.
• Data governance. Client confidentiality doesn’t pause because a tool is involved. Firms need clarity on where data is processed, who can access it, and whether it’s ever used to train a model.
• Adoption without structure. A tool rolled out without workflow integration, training, or internal policy tends to get used inconsistently — which is itself a governance risk.

‍

Closing the gap is a design choice, not an afterthought

This is exactly why Saga builds governance into the platform rather than treating it as a policy document nobody reads. Legal research across the platform is fully source-linked and cited, so every answer can be traced back to the underlying law, regulation, or case. Data handling runs on ISO 27001-certified infrastructure and in full GDPR compliance, with no client data ever used for training, end-to-end encryption, and regular independent security audits.

And because governance is a people problem as much as a technical one, Saga’s Amplify program exists specifically to make adoption governance-aware from day one — helping firms integrate AI into real workflows, train their teams, and build the internal clarity that regulators and bar associations are now asking for.

The firms that close the responsibility gap early won’t just be more compliant. They’ll be the ones clients and regulators trust first, as the rules catch up with the practice.

Want to see how a governed, source-linked AI platform looks in practice? Get a demo →