AI changes the playing field for intelligence analysis and background checks
- info381584
- 9 hours ago
- 4 min read
Look Closer is strengthening its capabilities in digital intelligence gathering and analysis by recruiting AI expert Magnus Rosell, who brings experience from FOI and Recorded Future. He sees great opportunities, but also enormous challenges, in the rapid technological advances in AI in recent years.
As a student in the late 1990s in the Engineering Physics programme at KTH Royal Institute of Technology, Magnus Rosell took a course in language technology, and was hooked. He continued his research at KTH (PhD in 2009) and, in his later professional roles, has closely followed the development of what we today call large language models and generative AI, capable of producing text, images, and video with abilities that in some areas surpass human intelligence.
"If I were to summarise what I’ve been doing all these years, I’d say that my entire professional career has revolved around automatic monitoring and analysis of unstructured information in one form or another. At KTH, I worked on algorithms to automatically provide an overview of the content in large volumes of news texts.”
As a developer at Recorded Future in 2011, he helped create the company’s event detection system that could draw conclusions about the future by analysing vast amounts of data, such as news sources and social media posts. Areas of focus included corporate acquisitions, military exercises, and political protests. Since then, Recorded Future has become a global company offering a range of threat intelligence services used by major corporations and government agencies in numerous countries.
At FOI, where he worked as a researcher between 2013 and 2023, Magnus Rosell studied what he calls semi-automatic intelligence monitoring. An important aspect of this was developing methods to detect influence operations, including systems to identify fake images and false posts on social media. He believes such systems, helping individuals and societies to detect, track, and understand AI-generated material, will become increasingly important. He describes the struggle between those who use AI for malicious purposes and those who seek to defend against it as similar to the long-running battle between virus and antivirus developers.
”The balance of power between the two sides will shift over time. In general, I believe that some form of stamp or watermark for AI-generated content will have to be developed. But the AI platforms themselves must also improve their traceability and source linking, so that underlying facts can be scrutinised and the creator’s intentions understood.”
Better background checks with AI support
At Look Closer, Magnus Rosell is looking forward to applying his theoretical knowledge in new, practical contexts. Concretely, this means using AI-based support for in-depth background checks and investigations, where the company’s researchers and analysts can receive assistance throughout the project, from information gathering to analysis, reporting, and client advisory.
”It’s an exciting challenge to help develop practical applications using AI, an area where revolutionary developments are truly taking place. Background checks are particularly interesting since the data sources are so numerous and unstructured, and Look Closer maintains such high standards of quality. My hope is that we can speed up certain processes, but of course also improve the final product.”
He does not envision a general AI for background checks, but rather many different types of agents and systems that handle specific tasks, from simple conversion and proofreading to more advanced analysis. A key consideration is what types of data are appropriate to process with which automated methods, both ethically and legally. The legal landscape concerning data handling, AI, and background checks will continue to evolve in the coming years.
”Something the company is already discussing extensively is how to assess and weigh historical adverse matters when making a business decision or completing a recruitment process. One could imagine that AI might critically review conclusions, look for alternative explanations, and help ensure the report is as balanced as possible. It’s probably a long road ahead, but regardless of how much AI can contribute, the final judgement must always rest with the analyst.”
Whether or not AI tools are used, there are many potential sources of error, from mistakes in data collection to the simple fact that publicly available material may be incomplete. It becomes even more difficult when individuals deliberately insert false information or falsify entire CVs or company presentations to deceive a recruiter or buyer. Look Closer’s analysts specialise in detecting such shortcomings, and with new technology, this capability will only improve further.
”Look Closer’s work depends on having the right data input. AI methods, too, need the right data to deliver good results. That’s where I’ll be putting much of my focus initially. With access to the right data, and in collaboration with our experienced analysts, I believe we can both shorten lead times and raise the quality of parts of the process and the final product with the help of AI.”
Many future “aha” moments
Throughout his career, Magnus Rosell has experienced many “aha moments” when new AI breakthroughs have redrawn the map of what is possible. He mentions, among others, the insights he gained from reading about Google’s language model BERT, the solution of the protein-folding problem through AlphaFold, which led to a Nobel Prize in Chemistry in 2024, and today’s multimodal AI models.
”A multimodal model can handle text, sound, images, video, and other types of data simultaneously. This makes it, for example, an excellent core for a robot, since it can address one of the most difficult problems in robotics, the many planning tasks a robot faces. These models are starting to manage that.”
Magnus Rosell is convinced that the future will bring many more “aha moments”. Predicting what will happen – and when – is, however, more difficult.
”I see many potential paths forward, but I don’t know which ones will be realised. One thing I’m very hopeful about is automatic research, which is already happening to some extent. If we can multiply our human research capacity several times over with the help of AI – what will the consequences be? The outcome is impossible to foresee, but it’s easy to imagine that it will have enormous effects. Even if AI and language model capabilities were to remain where they are today, it would still take a very long time before we have fully exploited their potential. They can be used in so many ways that I’m convinced we’ve only just begun.”
Comments