(Editor’s Note: This article is part of the Just Security symposium “Thinking Beyond Risks: Tech and Atrocity Prevention,” organized with the Programme on International Peace and Security (IPS) at the Oxford Institute for Ethics, Law and Armed Conflict. Readers can find here an introduction and other articles in the series as they are published.)
The tools available to human rights researchers have expanded dramatically over the past 20 years, enabling greater remote investigative powers than ever before. Analysts in distant locations working independently, in loose collectives or for formal NGOs, can now parse social media feeds, analyze satellite imagery, and examine geographical data that were once the preserve of government intelligence agencies.
As these technologies have become more readily available to a wider variety of actors, funders and governments have increasingly directed resources toward open-source investigation (OSINT) efforts, which can be launched rapidly as crises unfold and redeployed as situations change. Yet, this focus often comes at the expense of building local community networks that can provide a more varied dataset gained from proximity, lived experience, and local knowledge. Whereas OSINT efforts can be stood up immediately, such networks must be developed well before peak information demand. This process requires longer lead times and sustained financial and personnel resourcing that often stretches beyond the short-term (and frequently reactive) institutional funding timelines for crisis response.
Despite OSINT’s potential, the vast and ever-growing volume of open-source data and advanced information-gathering tools available today have yet to translate into real-world gains for atrocity prevention. The challenge is partly one of scale. As Human Rights Watch Washington Director Sarah Yager noted in 2023, “The sheer amount of information out there makes spotting a human rights problem solely through open-source intelligence a Sisyphean task.”
It is also one of capacity and process. Human rights activists from Sudan and Myanmar have testified to the U.K. Parliament that the government could – and should – have foreseen the atrocities unfolding in their countries if officials had engaged with a broader range of communities and pieced together disparate available data streams. However, the financial, personnel, and analytical resources required to collect, process, and interpret this data were lacking. As the UK Atrocity Prevention Working Group has noted, the problem is not that governments and organizations don’t have enough data, but that they fail to recognize, aggregate, analyze, and integrate it effectively into policymaking.
By definition, OSINT is readily accessible and publicly available information – but what enters the public domain remains shaped by existing power, privilege, and capacity dynamics. While today’s information environment is more democratic than ever, OSINT alone rarely provides the actionable information needed for effective atrocity prevention. Institutional funders should acknowledge this shortcoming and invest in a more diverse, multi-pronged approach to data gathering – whether by supporting partnerships or fostering coalitions – ideally substantively led by affected community members. Doing so would enable the timely and informed policy action urgently needed.
OSINT Isn’t Neutral
The Berkeley Protocol on Digital Open-Source Investigations defines OSINT as “publicly available information that any member of the public can observe, purchase or request without requiring special legal status or unauthorized access.” Specifically, it defines digital open-source information as publicly available information in digital format, including content posted on social media platforms, documents, images, videos and audio recordings shared online, satellite imagery, and government-published data.
A key reason why OSINT alone falls short as an early warning information source is that it’s not a neutral dataset. About one-third of the world’s population still lacks internet access , with the majority residing in South and East Asia and Africa – regions where many countries face an increased risk of mass killing. These areas are often virtually absent from the digital map, limiting the availability of detailed street-level imagery and making it harder for OSINT investigators to verify videos, photos, or satellite imagery. For instance, when researchers at UC Berkeley’s Human Rights Center sought to corroborate the location of alleged abuses for a report on gender persecution in Iran, they struggled with limited street-level imagery and ultimately relied on online images searches. Compounding this challenge are disparities in internet usability for speakers of lesser-known or historically underrepresented languages. In a world of more than 7,000 languages, more than 75 percent of users access the internet in just 10 languages, further limiting the accessibility of OSINT data and tools.
Another major limitation of OSINT for early warning is that media and information environments vary widely across the world, with the ability to freely share information depending on basic rights and infrastructure that are not universal. The stark contrast between the abundance of data on Ukraine compared with the paucity of information on Ethiopia underscores this asymmetry – while some contexts have an abundance of open-source data, while others remain severely under-documented.
This disparity has real-world policy consequences. For example, in testimony before the U.K. Parliament, Sudanese activists noted as early as summer 2023 that the looming crisis in El Geneina in West Darfur was being overlooked due to a lack of internet and telephone infrastructure, whereas events in the capital, Khartoum, received far greater attention and acknowledgment.
Counterbalancing OSINT’s Asymmetry
The uneven availability of open-source information means it must be combined with other documentation strategies, tools, and datasets to serve effectively as an early warning alarm. Institutional funders, OSINT investigators, and tool developers should take proactive steps to address these gaps by expanding data redundancy, prioritizing local context, and promoting equitable partnerships.
The following four concrete actions – two for international OSINT investigators, one for software and tool developers, and one for funders – can help counterbalance OSINT’s inherent asymmetry:
International OSINT Investigators: Build in Redundancy
To maximize the effectiveness of open-source intelligence, international OSINT investigators should build redundancy into their documentation strategies by drawing from multiple, overlapping sources of information. There are more ways than ever to document and collect data on human rights violations, making it easier to generate and share critical information. From social media posts by eyewitnesses or perpetrators to satellite imagery, open-source researchers can gather information from a range of sources and tools.
This collection of tools, or “documentation stack,” is similar to the tech sector’s concept of a “tech stack” – a suite of tools, platforms, software, and apps a company uses to carry out its operations and track its performance. A documentation stack consists of resources, tools, software, and platforms that analysts and organizations use to gather data on human rights abuses. With the advent of social media and decreasing costs of technology, this stack has expanded to include testimony, physical documents, videos, images, satellite imagery, surveys, reporting, geolocation tools, sensors, and physical and digital forensic tools. The more tools included in the stack, the greater the redundancy within the human rights information-gathering ecosystem – making it harder to suppress, distort, or ignore concerns about potential atrocities.
For example, if a government shuts down the internet around an election, local monitoring networks can still document events on the ground, while satellite imagery can produce real-time visual records. If satellite imagery is too costly or lacks the necessary degree of granularity, surveys and interviews conducted by journalists, humanitarian actors, or community groups can help fill in information gaps, providing insights into displacement patterns or rising ethnic tensions.
Beyond closing data gaps, redundancy strengthens the resilience and credibility of open-source information, making it harder for detractors to undermine the truth. By combining multiple data streams, OSINT analysts and organizations can fortify information against attack, denial, or competing narratives. The more tools available to verify information, the greater the weight of that information and analyses that rely on it.
For instance, to investigate an April 2022 cluster munitions attack on the Kramatorsk rail station in eastern Ukraine, visual investigations practice SITU Research and Human Rights Watch combined witness testimony, data on munition performance, satellite imagery, and video documentation. The assault had killed at least 58 people and injured more than 100, many of them civilians. The resulting findings helped generate evidence contradicting Russia’s denial of responsibility and indicating possible war crimes.
International OSINT Investigators: Prioritize Local Context
In addition to building in redundancy, international OSINT investigators should elevate local context in both data collection and analysis. OSINT enables analysts – whether working independently or as a part of formal or informal collectives – o gather individual pieces of data, much like gathering scattered puzzle pieces. However, for these individual data points form a meaningful picture, they must be connected and contextualized. This is a strength of the documentation stack: combining diverse types of data, analysis, and sources to provide greater clarity and insight.
Yet analyzing gathered information, particularly for early warning signs, is no simple feat. While open-source investigations are often imagined as conducted far from the ground, effectively sifting through data gathered requires deep contextual knowledge, including of local languages, subtext, jargon, and cultural shorthand. Often, small and seemingly insignificant events can signal larger atrocities on the horizon. Understanding which small signals are important or insignificant requires a robust grasp of local context and situational dynamics. This is particularly true in situations where analysts must detect and interpret small, seemingly unconnected indicators rather than a single violent act or specific pattern of events.
Just as foreign correspondents have long relied on the situational awareness of local journalists to enhance the substance and accuracy of their reporting, OSINT teams should incorporate insights from affected communities and others with relevant lived experience. This can take several forms, including hiring community members on international OSINT teams, building collaborative partnerships, or providing OSINT training and support to existing local NGO or activist networks. While a growing number of multi-disciplinary OSINT teams – such as the Yale School of Public Health’s Humanitarian Research Lab – are hiring analysts with lived experience or deep knowledge of local dynamics, full and reciprocal partnerships between locally-led groups and internationally-based OSINT teams are the exception rather than the norm.
At the same time, international OSINT practitioners must take care not to reinforce power dynamics between themselves and local sources. Respecting and prioritizing local knowledge not only saves valuable time and resources – which would otherwise be spent on training outsiders to understand the context – but is also as mission-critical as technical OSINT expertise. As such, the provision of local knowledge and information should be compensated with appropriate credit (as security conditions allow) and funding. This happens far too infrequently, and should become a standard practice.
Social Media Companies and Tool Developers: Create Equitable Data Access Policies
When it comes to the private sector, social media companies and tool developers should create more equitable data access policies and pricing structures accessible to the full range of OSINT researchers and analysts. Previously, a variety of tools were easily accessible to mine data from social media platforms and the internet. For instance, Application Programming Interfaces (APIs) enabled researchers globally to analyze social media content to corroborate allegations of human rights abuses and track atrocity early warning signs.
However, access to APIs has become either increasingly expensive and narrow, or restricted altogether – such as X’s API, which went from being freely available to exorbitantly costly in April 2023, or Facebook’s CrowdTangle, which was shut down in August 2024. Of those APIs that remain available, many are either paywalled or geographically restricted – for instance, access to TikTok’s API is limited to researchers located in the United States and Europe.
These deepening restrictions pose a significant entry barrier for OSINT investigative teams, independent non-American or European analysts, and locally-led initiatives in volatile regions. As Thraets Research Lead Richard Ngamita notes, such barriers make “…it exponentially difficult for global south researchers to study misinformation and disinformation crises,” locking out many independent actors in the global south who rely on open source data to monitor and analyze social media developments.
Funders: Incentivize Cross-Organizational Collaboration
Lastly, international funders – including governments, international organizations, foundations, and private donors – should incentivize collaboration across organizations and sectors. Strengthening OSINT analysis through diverse data streams, elevating local knowledge without being extractive, securing API access, and advocating for fairer API policies all require significant resources. However, not all OSINT teams can excel in all these areas. Given the constrained funding landscape for human rights investigations and atrocity prevention, organizations often stretch beyond core capabilities to secure funding – an approach that is neither sustainable nor efficient in the long run.
Many organizations and teams specialize in specific aspects of the multi-pronged documentation and investigative strategy proposed. To harness these strengths, funders should incentivize collaborations that leverage each organization’s unique expertise. Supporting partnerships among analyst teams and organizations can foster sustained, multi-pronged investigative efforts that merge the distinct capabilities of various OSINT actors. While such collaborations are rare, they do exist—examples include the Investigative Commons, a Berlin-based collective of investigators, reporters, lawyers, and artists collaborating on targeted investigations, and Amnesty International’s Digital Verification Corps, a network of universities partnering with Amnesty to collect and verify digital content.
Crucially, funders and organizations often underestimate the complexities of integrating OSINT expertise across geographies. Successful collaborations require clear, reciprocal benefits, a shared mission, and strong trust among participants, along with dedicated management capacity. These elements are not developed overnight, and require consistent, thoughtful engagement among collaborators. To facilitate this, funders should allocate specific resources for grantees to explore, establish, and strengthen collaborative efforts.
Conclusion
While new and emerging open-source tools offer exciting possibilities, OSINT investigative teams should integrate them as part of broader, multi-pronged documentation stack – one that is supported at all levels by local knowledge and insights. To effectively identify early warning signs of atrocities, international OSINT teams must incorporate situational expertise that helps decode complex, often-hidden indicators of escalating violence.
More broadly, the atrocity investigation ecosystem must shift toward a more sustainable, collaborative model that prioritizes local expertise and cross-organizational cooperation. Strengthening such partnerships will lead to more comprehensive early warning and documentation efforts — something funders should recognize and prioritize. As the funding landscape become increasingly constrained and atrocity dynamics grow ever more complex, harnessing collective expertise, diverse sources, and integrated data streams offers the best path forward toward building a stronger, more effective system for investigating and documenting the most serious international crimes.