MAIA: Artificial Intelligence to Protect Those Who Defend Rights

In Latin America and the Caribbean, civil society organizations face an increasingly complex landscape. Repression no longer occurs only in the streets or in courtrooms; it also unfolds in digital spaces. Mass surveillance, targeted cyberattacks, disinformation campaigns, gender-based digital violence, communication infiltration, and online criminalization are now part of the daily reality for human rights and land defenders, social organizations, unions, community media outlets, and many other individuals and groups at risk.

In response to this context, MAIA emerges: AI-Powered Threat Modeling, a tool designed from and for civil society organizations in Latin America and the Caribbean. MAIA is not just a technological innovation; it is a political and ethical commitment to building safer, more sovereign, and more just digital territories.

Social organizations face complex digital threats, yet they often lack the technical, financial, and human resources to manage them effectively. Although robust methodologies such as threat modeling are widely used in the fields of digital and operational security, their implementation requires specialized knowledge and intensive manual work.

Threat modeling involves identifying valuable assets (information, communications, identities), potential adversaries, types of attacks, probabilities, and impacts. It is a rigorous and effective process for prioritizing risks and making strategic decisions. However, in practice, it is often inaccessible to organizations with limited budgets or those facing urgent risk situations.This creates a critical gap: the tools exist, but they are out of reach for those who need them most.

MAIA was created to close that gap.

The project proposes developing a specialized Small Language Model (SLM) for threat modeling, capable of reducing between 90% and 95% of the manual work required in these processes. This means tasks that previously took hours or even days can be completed in a fraction of the time, allowing organizations to respond more quickly in high-risk contexts.
But MAIA is not simply “AI applied to digital security.” Its design is grounded in three fundamental principles:

1. Technological Sovereignty and Data Protection
The information handled by civil society organizations is highly sensitive. Using commercial platforms that process data on servers located outside the region and under foreign jurisdictions entails unacceptable risks.For this reason, MAIA is built on:
* Open-source language models
* Fully local deployment
* Infrastructure hosted within Latin American territory
* End-to-end encrypted communications
* Architectures that prevent external access to information

MAIA demonstrates that it is possible to develop powerful artificial intelligence without sacrificing privacy or autonomy.

2. Intersectional Co-Creation

MAIA is not built “for” communities, but with them. The project incorporates a co-creation process with Indigenous organizations, unions, feminist groups, territorial defenders, and alternative media outlets. These organizations are not passive users; they are co-researchers and co-designers.This approach recognizes something fundamental: those who experience digital repression hold expert knowledge that must guide technological design. Incorporating community knowledge, lived experience, and decolonial perspectives is not an add-on—it is at the heart of the project.

3. Democratizing Knowledge in Digital Security

Until now, threat modeling has been primarily accessible through in-person workshops facilitated by specialists. This limits its reach and sustainability.MAIA transforms this reality by turning a complex methodology into an accessible, guided, and interactive tool. This will enable resource-constrained organizations to:
* Identify priority digital risks
* Assess impacts and probabilities
* Design concrete countermeasures
* Integrate digital security into their everyday strategic planningIn other words, MAIA turns digital security into an integrated and ongoing practice, rather than a one-time intervention. Speaking of “digital territories” means recognizing that virtual space is not neutral. It is a field of political, economic, and cultural dispute.
Just as communities defend physical territories against extractivism, they must also defend their digital territories against surveillance, criminalization, and technology-enabled violence. The tool does not seek to replace human judgment. On the contrary, it functions as an assistant that strengthens the strategic capacity of organizations, always keeping supervision and decision-making in human hands.

The project explicitly acknowledges the risks associated with artificial intelligence: algorithmic bias, misinformation or model errors, potential misuse by hostile actors, environmental impact, and excessive dependence on the tool. This approach seeks not only to mitigate risks, but also to set a precedent for how responsible technology can be developed in contexts of high political sensitivity.

MAIA is not merely a digital security tool. It is also a statement: artificial intelligence can be developed differently.It can be local rather than dependent, collective rather than corporate, ethical rather than extractive, sovereign rather than subordinate.By strengthening organizations that defend labor rights, land rights, gender justice, and Indigenous rights, MAIA indirectly contributes to the defense of democracy and social justice across the region.The expected impact of MAIA is deeply transformative. By drastically reducing the technical burden of risk analysis, access to self-protection tools expands. This enables more organizations to integrate digital care practices into their daily routines, strategic decisions, and organizational processes. In the long term, the project seeks to weave a culture of security in which technology ceases to be an instrument of control and becomes an ally for collective emancipation.At a historical moment when digital repression is expanding and artificial intelligence is increasingly concentrated in the hands of large corporations, MAIA represents a concrete alternative: AI developed from Latin America, for Latin America, in service of those who defend rights and territories.Because building secure digital territories is not only a technical matter. It is a condition for struggles for justice, dignity, and autonomy to continue.