Skip down to main content

Open Source Intelligence (OSINT) and AI: The Informational Pivot of Intelligence Analysis

Open Source Intelligence (OSINT) and AI: The Informational Pivot of Intelligence Analysis

Published on
17 Feb 2023
Written by
Mariarosaria Taddeo, Luciano Floridi and Riccardo Ghioni
In their new blog, Riccardo Ghioni, Mariarosaria Taddeo, Luciano Floridi, share their insights on the growth of OSINT and AI, and reflect on the opportunities and challenges for governments, defence and civilian organisations.

Open Source Intelligence (OSINT) and AI: The Informational Pivot of Intelligence Analysis

By Riccardo Ghioni, Mariarosaria Taddeo, and Luciano Floridi

In their new blog, Research Fellow Riccardo Ghioni, University of Bologna, Associate Professor and Senior Research Fellow, Mariarosaria Taddeo, Oxford Internet Institute and Professor Luciano Floridi, Oxford Internet Institute share their insights on the growth of OSINT and reflect on the opportunities and challenges for governments, defence and civilian organisations.  The academics also highlight the importance of effective governance and oversight to ensure the future reliability of intelligence practice.

Most data online is open source, meaning interested parties can freely access them with relatively low technical requirements. In the right setting, such data can become open source intelligence (OSINT), i.e. publicly available information exploited for intelligence purposes. Over the years, OSINT has become an integral part of intelligence practice, with technological progress delivering new collection methods and creating new intelligence sources, such as satellite images, social media, public records, and digital currencies. Indeed, many estimates place OSINT at around eighty per cent of all the intelligence material used by law enforcement agencies. Recently, there have been many prominent cases of digital OSINT used as valuable intelligence by both state agencies and activist networks.   For example, in 2018, a group of OSINT analysts was able to geolocate the execution site featured in one of the most famous Islamic State propaganda videos. This was achieved by comparing the terrain, vegetation and building characteristics shown in the video frames against satellite images from Google Earth.

More recently, the use of OSINT in the ongoing Russo-Ukrainian conflict has impacted contemporary warfare by outsourcing parts of the targeting cycle to civilians. The widespread use of smartphones among citizens has enabled military personnel to exploit intelligence gathered by the civilian population and shared on social media to obtain approximate location estimates for enemy combatants. This has raised significant concerns about the extent of military surveillance of civil society, and the risks of involving the civilian population in military operations.

As with many other aspects of contemporary society (and perhaps even more so), Artificial intelligence (AI) has affected the retrieval and analysis of OSINT data.  As computational power becomes cheaper and algorithms more sophisticated, more data can be acquired and processed in almost real-time. Moreover, AI can also assist the analysis phase of the OSINT cycle, generating valuable intelligence based on pre-trained models and thus countering the information overload problem faced by intelligence analysts. For instance, many hours of video footage can be automatically scanned and labelled for future use, while long documents can be translated and summarised by AI algorithms. Social media can be mined to assess opinions about a given topic or to geolocate a particular person based on the content they post. Queries can then be run on a database, and the relevant intelligence is returned to the analyst. As AI becomes more integrated into everyday OSINT analysis, it becomes clear that those who can train state-of-the-art AI algorithms can fully exploit the goldmine of publicly available data and produce better intelligence products. In a 1904 article titled The Geographical Pivot of History, geographer Halford John Makinder laid down the foundations of geopolitical theory as a struggle over the control of the Heartland, a strategically dominant and resource-rich area roughly corresponding to contemporary Russia. Extending this analysis to the intelligence domain, AI becomes the informational pivot through which the timely collection and analysis of open-source intelligence is made possible and provides an information advantage to the analyst. Paraphrasing Makinder’s famous quote, who rules AI commands intelligence, thus breaking the monopoly of intelligence collection and analysis held by intelligence agencies and projecting it into the social dimension.

In our recently published article, we discuss the role AI has played and will play in the context of OSINT and focus on the emerging area of research known as the Governance, Ethical, Legal and Social Implications (GELSI) of OSINT. Indeed, we are the first to reference and address this GELSI framework in an academic paper. We do so by systematically searching for GELSI-themed articles within the relevant literature and providing a thorough review of the existing scholarship. As it turns out, current concerns focus on the micro (individual) and macro (societal) issues around adopting AI algorithms for OSINT analysis. At the micro-level, we find privacy issues relating to how user data is handled during investigations, considering anonymisation and increased regulation as potential solutions to privacy breaches. Some attention is also given to the development of privacy-friendly software, which would restrict the ability to link citizens to their online personas. At the macro level, the focus is on the role of OSINT in shaping society as a whole. The emergence of activist networks and other non-state actors as key players in the OSINT landscape brings up two related phenomena. First, the democratising force of open-source information, which increases the potential for citizen oversight of government activities. Second, the hidden threat of misinformation disguised as grassroots activism. Indeed, malicious actors may try to poison the well of OSINT analysis by deliberately providing misleading information, which could then lead to doxing of innocent citizens. Thus, the need to verify sources and cross-reference key data points becomes essential, even more so as intelligence collection becomes automated and risks including potentially inaccurate material. Finally, as we have hinted above, integrating AI tools into OSINT analysis creates an asymmetric advantage which favours the actors with easy access to more advanced algorithms, richer datasets, and computational power. These asymmetries are likely to manifest between the public and private sectors and between state and non-state actors. In these conflicts, those who can exploit the informational pivot better will prevail and determine the future of intelligence practice.

It is hard to tell which transformation will have the heaviest impact on OSINT practice. However, as more phases of the intelligence cycle become automated, it is clear that a suitable framework for validating AI-powered OSINT must be developed to address reliability, transparency and oversight issues. If not, the opacity of traditional intelligence will become even worse, and it will be hard to disentangle facts from misinformation and unilaterally determine responsibility for intelligence decisions.

Download their full paper; ‘Open source intelligence and AI: a systematic review of the GELSI literature’, published in AI & Society.

About the authors

Related Topics