On behalf of the Attorney General I was assigned to review all classified all source documents concerning Nazi war crimes committed in Belarus. My impression is that 80 percent of classified information can be obtained from open sources but one must have professional advice on where to look. Later I was required to submit my book manuscripts to CIA pre-publication review. I frequently needed open sources to justify writing about still classified topics. I had the great good fortune to have several mentors in the intelligence community who pointed me towards obscure books and archives. Lexus/Nexus digital archives of newspaper and magazines are almost as good as FBIS but quite expensive. Law students, however, have free access to this search service.
Where does the illustration come from? The types of information collection in it are textual, as you describe in the article, FBIS, news reports, etc. . But that's not the current "OSINT" journalists get sent to training for these days --- for identifying locations from IPhone photos and identification of weapons from serial numbers -- "Visual Information."
There seem to be more and more definitions of OSINT these days. I assume the intelligence community was the originator. It seems to have gotten away from them.
With the pervasiveness of agents of undetermined loyalty, open sources may be the best way to test them and their information. However open sources can also be spoofed. It takes work to do good open source collection. And lots of sources, processing and even thought.
I think the intersection of AI search engines and digital archives makes vetting sources a bit easier. In the old days we had to use Boolean search terminology. A pain.
Back in the days before the war in Vietnam and the resulting unpopularity of a military presence on college campuses, many of the Army Reserve's approximately 60 Military Detachments (Strategic), aka MID(s) were intentionally located in cities where there were also major universities. In some cases, those MID(S) actually conducted their weekend drills on campus where they could access the facilities of the university's library. The war, unfortunately, brought an end to many of those relationships.
The problem many of the Reserve units had was that we could not store any material classified above the SECRET level. So that was all we had to work with when we were at home station. We had to wait until we went on active duty and could also access TS/SI material by working in the SCIF of whatever Active unit we were supporting. Or of another SCIF closer to where we were stationed, if one was available.
When the MID(S) I commanded (469th, Ft Snelling, St. Paul, MN) was lucky enough to be able to use a Navy Reserve SCIF just down the road from our Reserve Center, that gave us a signficant advantage. We were then shifted to support the European Command and its Joint Analysis Center, located at RAF Molesworth, UK. Good thing we had access to that Navy Reserve SCIF! We were the first MID(S) to establish direct secure computer links to the active duty unit we supported.
Since then, a number of large SCIFS have been constructed so that Reservists can more readily access classified material at the TS/SI level. That also gave us the ability to more easily access OSINT via computers, with the advantages (and disadvantages) pointed out above.
It's good to remember, even with good access to TS/SI material without going to England, that we should not forget--thanks to the Internet--that we also have easier access to OSINT. Those of us in the Reserves made the transition to the brave new world of the 21st century late in the 20th century. Now all we have to remember, with the examples of an airman sharing highly classified data with his gamer buddies and another intelligence analyst attempting to impress his Internet girlfriend with his access is that the weak link in security is people. The wrong person with access to a SCIF can defeat the very best secure computer systems.
I was hired by FBIS in 1992, and my last day with OSE was 30 September 2018, following OSE's decision to shut down overseas operations. OSE has experienced many pendulums swinging back and forth regarding its mission and place within the wider IC community, so what's said in the article is nothing new. A relatively new thing is that only US citizens can work on OSINT, thus making it a bona fide US intelligence discipline and operation. FBIS/OSC (Open Source Center)/OSE was unique back when the CIA regarded it as an inferior discipline because of the number of foreign linguists in its many bureaus worldwide. To US staffers, the main attraction of working for FBIS/OSC/OSE was the possibility of spending a few years in one of its many bureaus worldwide. We began monitoring online sources well before 9/11 and social media to identify foreign fighters in Syria even before ISIS became a household name. Our findings were of vital importance. Yet, the bureaus and their employees were deemed a security risk, probably a justifiable concern considering the shifts in political sentiments and alliances and cyber threats that have grown exponentially in the last decade. The situation post-2018, from my experience, is the OSE does not do the work alone; it outsources OSINT work. Contractors tend to neglect the linguistic and cultural aspects of OSINT and focus on technology, thus turning OSINT into a big data collection operation. In my opinion, substantive linguistic and cultural knowledge and continuity in the job are some of the key elements of intelligence gathering. In my experience, the US linguists who work for OSE currently are second or third-generation Americans who have basic linguistic knowledge but lack cultural expertise, and no continuity as they are encouraged to advance their careers by changing roles every three or four years and relying 100 percent on technology. We had many quirky characters, as described in the article, people who were intellectuals, spoke many languages, experienced Soviet persecution, fled their undemocratic countries, and regarded their work for Americans as a badge of honor. That's not the case anymore; working for Americans has become a security risk, too. What I'm trying to say is that OSE should tap I was hired by FBIS in 1992, and my last day with OSE was 30 September 2018, following the OSE's decision to shut down overseas operations. OSE has experienced many pendulums swinging back and forth regarding its mission and place within the wider IC community, so what's said in the article is nothing new. A relatively new thing is that only US citizens can work on OSINT, thus making it a bona fide US intelligence discipline and operation. Yet, what made FBIS/OSC/OSE unique back at the time when the CIA regarded it as an inferior discipline was the number of foreign linguists in its many bureaus throughout the world. The main attraction of working for OSE was the possibility of spending a few years in one of its many bureaus worldwide, so the US staff that came out to work in the field were also talented and open-minded. We began monitoring online sources well before 9/11 and social media to identify foreign fighters in Syria even before ISIS became a household name. Our findings were of vital importance. Yet, the bureaus and their employees were deemed a security risk, probably a justifiable concern considering the shifts in political sentiments and alliances and cyber threats that have grown exponentially in the last decade. The situation post-2018, from my experience, is the OSE does not do the work alone; it outsources OSINT work as it lacks capabilities. Contractors tend to neglect the linguistic and cultural aspects of OSINT and focus on technology, thus turning OSINT into a big data collection operation. In my opinion, continuity is a key element of intelligence gathering. You don't get continuity by changing strategies and encouraging employees to advance their careers by changing roles every three years and relying 100 percent on technology. It's cheaper, but not better. We had many quirky characters, as described in the article, people who were intellectuals, spoke languages, fled from pro-Soviet governments, and regarded their work for Americans as a badge of honor. This is not the case anymore; working for Americans has become a security risk, too. So, good luck, OSE, in your future endeavors.
Interposing AI between a researcher and the sources may be a huge mistake. LLM AIs are inherently untrustworthy since they are trained by scraping the internet including all the garbage that is on it. They frequently hallucinate. GIGO is always a risk. Just look at the lawyers caught out by using AI to research case law. Custom-built algorithms are also prone to bias so, again, humans have the edge. Modern-day OSINT is not so much text-based as analysing databases, video, audio, and still images. Especially metadata. I foresee a long future for flesh-and-blood researchers who eschew AI and outperform those who don't.
Bellingcat has a good piece on the bad use of OSINT which I am sure will resonate with many intel pros:
I apologize for my slow response to Henry R. Schlesinger’s excellent article on open-source intelligence (OSINT). Any superior analyst I have ever worked with or met depended heavily on OSINT to develop an understanding of the region or target they were assigned to cover. Yet too many observers today comment on OSINT as if it was the next big new thing.
Over the years I have witnessed several attempts at improving the amount of OSINT available to analysts. The only one that ever worked well and consistently for me was FBIS. Not surprisingly the Intelligence Community (IC) in all its wisdom decided to end this valuable source of information I think around 1996. At some point in the 1990s the Directorate of Intelligence (DI) formed an organization [I’ve forgotten its name] that not only produced excellent analysis strictly from OSINT, but its reports often proved more insightful than those of their DI counterparts who had the advantage of having access to classified information on the topic. As best I can tell the organization I examined no longer exists.
A big part of the difference between the two analytical corps was that the OSINT group, except for the DI Division Chief, was not cleared for access to classified information, and for a very good reason. Most, if not all, of its analysts could not have passed the standard qualification test to receive a clearance. Virtually all had friends or relatives they were still in contact with overseas. On the plus side, they also differed in that virtually all were native speakers of Chinese, Korean, Arabic, Persian, or Russian, with the majority having received a PhD degree from a top US institution. Also, they were not focused primarily current reporting, but research projects on our principle adversaries.
I came to know this group from a project on North Korea requested by the DNI’s manager for North Korean issues at the time, Ambassador Joe DeTrani. It came about in the aftermath of the DIA underestimating the size of the North Korean ground forces by at least a third. My assignment was to determine how much research had been done prior to our discovering the mistake. My methodology consisted of reading every report in the previous ten years produced by anyone that referenced North Korea’s ground forces. My first discovery was I couldn’t find a single true research project. Several reports were described as research, but upon closer examination they all proved to be simple summaries of current intelligence reports. With this mind, I pursued on my own a similar exercise on WMD research papers produced from the Gulf War and the time of the Iraq WMD fiasco only to find that like North Korea that in this case there had been no all-source in-depth research focused on the question of Iraq’s WMD program during this critical period.
Having examined all reports published on North Korea I recognized that the OSINT analysts also often provided more insightful current reporting than the DI was producing. When I sought to dig a bit deeper I discovered that the DI analysts were jealous of the OSINT analysts, and rarely took their views into account in preparing their own reports. I found that this could be explained in large part by many all-source analysts’ preference for classified information, but in a number of cases it would also have required the DI to admit that several of its understanding of North Korean military behavior were likely incorrect. An understandable reaction maybe, but in my own experience as an analyst I found that some of my most important discovers were when I realized that my hypothesis was wrong, and I needed to rethink the puzzle I was attempting to solve.
Thinking back over my time as an analyst at DIA and later CIA, the biggest problem I encountered with using OSINT was the difficulty in asking the main collection organization, most often FBIS, a specific question or for it to put a special emphasis on a particular topic or look back at old data. To do this, I had to go through a large an complicated separate bureaucracy that most often proved to be not worth the effort. Equally dauting was the requirement of sorting through by hand myself all the mentions of a particular word or issue I was investigating. My solution in one particular case was to adopt the WWII punch card system where I could record every mention made of China’s militia, punch out each specific variable(s) mentioned and make a note on the card of my assessment of its relevance and importance to my research topic. Since no one is doing in-depth research anymore I expect few run into this problem today.
While I applaud CIA taking on a greater OSINT responsibility, I think we should broaden our I thinking. I would like to see a brand-new collection agency formed comparable to NSA or NGA exclusively for the purpose of collecting open-source information similar to what FBIS did, but also include information obtained from the internet’s social media. Much like single-source agencies such as, NSA and NGA, the new agency would produce current reports, but also be able to respond to individual research requests from all-source analysts. I think this is best done by collocating a few single-source OSINT analysts to advise and assist the all-source analysts at CIA, DIA, and INR. I know that this type of colocation worked especially well when I was at INR.
On behalf of the Attorney General I was assigned to review all classified all source documents concerning Nazi war crimes committed in Belarus. My impression is that 80 percent of classified information can be obtained from open sources but one must have professional advice on where to look. Later I was required to submit my book manuscripts to CIA pre-publication review. I frequently needed open sources to justify writing about still classified topics. I had the great good fortune to have several mentors in the intelligence community who pointed me towards obscure books and archives. Lexus/Nexus digital archives of newspaper and magazines are almost as good as FBIS but quite expensive. Law students, however, have free access to this search service.
Atty. John J. Loftus (retired)
Where does the illustration come from? The types of information collection in it are textual, as you describe in the article, FBIS, news reports, etc. . But that's not the current "OSINT" journalists get sent to training for these days --- for identifying locations from IPhone photos and identification of weapons from serial numbers -- "Visual Information."
There seem to be more and more definitions of OSINT these days. I assume the intelligence community was the originator. It seems to have gotten away from them.
.
With the pervasiveness of agents of undetermined loyalty, open sources may be the best way to test them and their information. However open sources can also be spoofed. It takes work to do good open source collection. And lots of sources, processing and even thought.
I think the intersection of AI search engines and digital archives makes vetting sources a bit easier. In the old days we had to use Boolean search terminology. A pain.
Back in the days before the war in Vietnam and the resulting unpopularity of a military presence on college campuses, many of the Army Reserve's approximately 60 Military Detachments (Strategic), aka MID(s) were intentionally located in cities where there were also major universities. In some cases, those MID(S) actually conducted their weekend drills on campus where they could access the facilities of the university's library. The war, unfortunately, brought an end to many of those relationships.
The problem many of the Reserve units had was that we could not store any material classified above the SECRET level. So that was all we had to work with when we were at home station. We had to wait until we went on active duty and could also access TS/SI material by working in the SCIF of whatever Active unit we were supporting. Or of another SCIF closer to where we were stationed, if one was available.
When the MID(S) I commanded (469th, Ft Snelling, St. Paul, MN) was lucky enough to be able to use a Navy Reserve SCIF just down the road from our Reserve Center, that gave us a signficant advantage. We were then shifted to support the European Command and its Joint Analysis Center, located at RAF Molesworth, UK. Good thing we had access to that Navy Reserve SCIF! We were the first MID(S) to establish direct secure computer links to the active duty unit we supported.
Since then, a number of large SCIFS have been constructed so that Reservists can more readily access classified material at the TS/SI level. That also gave us the ability to more easily access OSINT via computers, with the advantages (and disadvantages) pointed out above.
It's good to remember, even with good access to TS/SI material without going to England, that we should not forget--thanks to the Internet--that we also have easier access to OSINT. Those of us in the Reserves made the transition to the brave new world of the 21st century late in the 20th century. Now all we have to remember, with the examples of an airman sharing highly classified data with his gamer buddies and another intelligence analyst attempting to impress his Internet girlfriend with his access is that the weak link in security is people. The wrong person with access to a SCIF can defeat the very best secure computer systems.
Larry Brown COL (ret), MI-USAR
I was hired by FBIS in 1992, and my last day with OSE was 30 September 2018, following OSE's decision to shut down overseas operations. OSE has experienced many pendulums swinging back and forth regarding its mission and place within the wider IC community, so what's said in the article is nothing new. A relatively new thing is that only US citizens can work on OSINT, thus making it a bona fide US intelligence discipline and operation. FBIS/OSC (Open Source Center)/OSE was unique back when the CIA regarded it as an inferior discipline because of the number of foreign linguists in its many bureaus worldwide. To US staffers, the main attraction of working for FBIS/OSC/OSE was the possibility of spending a few years in one of its many bureaus worldwide. We began monitoring online sources well before 9/11 and social media to identify foreign fighters in Syria even before ISIS became a household name. Our findings were of vital importance. Yet, the bureaus and their employees were deemed a security risk, probably a justifiable concern considering the shifts in political sentiments and alliances and cyber threats that have grown exponentially in the last decade. The situation post-2018, from my experience, is the OSE does not do the work alone; it outsources OSINT work. Contractors tend to neglect the linguistic and cultural aspects of OSINT and focus on technology, thus turning OSINT into a big data collection operation. In my opinion, substantive linguistic and cultural knowledge and continuity in the job are some of the key elements of intelligence gathering. In my experience, the US linguists who work for OSE currently are second or third-generation Americans who have basic linguistic knowledge but lack cultural expertise, and no continuity as they are encouraged to advance their careers by changing roles every three or four years and relying 100 percent on technology. We had many quirky characters, as described in the article, people who were intellectuals, spoke many languages, experienced Soviet persecution, fled their undemocratic countries, and regarded their work for Americans as a badge of honor. That's not the case anymore; working for Americans has become a security risk, too. What I'm trying to say is that OSE should tap I was hired by FBIS in 1992, and my last day with OSE was 30 September 2018, following the OSE's decision to shut down overseas operations. OSE has experienced many pendulums swinging back and forth regarding its mission and place within the wider IC community, so what's said in the article is nothing new. A relatively new thing is that only US citizens can work on OSINT, thus making it a bona fide US intelligence discipline and operation. Yet, what made FBIS/OSC/OSE unique back at the time when the CIA regarded it as an inferior discipline was the number of foreign linguists in its many bureaus throughout the world. The main attraction of working for OSE was the possibility of spending a few years in one of its many bureaus worldwide, so the US staff that came out to work in the field were also talented and open-minded. We began monitoring online sources well before 9/11 and social media to identify foreign fighters in Syria even before ISIS became a household name. Our findings were of vital importance. Yet, the bureaus and their employees were deemed a security risk, probably a justifiable concern considering the shifts in political sentiments and alliances and cyber threats that have grown exponentially in the last decade. The situation post-2018, from my experience, is the OSE does not do the work alone; it outsources OSINT work as it lacks capabilities. Contractors tend to neglect the linguistic and cultural aspects of OSINT and focus on technology, thus turning OSINT into a big data collection operation. In my opinion, continuity is a key element of intelligence gathering. You don't get continuity by changing strategies and encouraging employees to advance their careers by changing roles every three years and relying 100 percent on technology. It's cheaper, but not better. We had many quirky characters, as described in the article, people who were intellectuals, spoke languages, fled from pro-Soviet governments, and regarded their work for Americans as a badge of honor. This is not the case anymore; working for Americans has become a security risk, too. So, good luck, OSE, in your future endeavors.
Interposing AI between a researcher and the sources may be a huge mistake. LLM AIs are inherently untrustworthy since they are trained by scraping the internet including all the garbage that is on it. They frequently hallucinate. GIGO is always a risk. Just look at the lawyers caught out by using AI to research case law. Custom-built algorithms are also prone to bias so, again, humans have the edge. Modern-day OSINT is not so much text-based as analysing databases, video, audio, and still images. Especially metadata. I foresee a long future for flesh-and-blood researchers who eschew AI and outperform those who don't.
Bellingcat has a good piece on the bad use of OSINT which I am sure will resonate with many intel pros:
https://www.bellingcat.com/resources/2024/04/25/oshit-seven-deadly-sins-of-bad-open-source-research/
I apologize for my slow response to Henry R. Schlesinger’s excellent article on open-source intelligence (OSINT). Any superior analyst I have ever worked with or met depended heavily on OSINT to develop an understanding of the region or target they were assigned to cover. Yet too many observers today comment on OSINT as if it was the next big new thing.
Over the years I have witnessed several attempts at improving the amount of OSINT available to analysts. The only one that ever worked well and consistently for me was FBIS. Not surprisingly the Intelligence Community (IC) in all its wisdom decided to end this valuable source of information I think around 1996. At some point in the 1990s the Directorate of Intelligence (DI) formed an organization [I’ve forgotten its name] that not only produced excellent analysis strictly from OSINT, but its reports often proved more insightful than those of their DI counterparts who had the advantage of having access to classified information on the topic. As best I can tell the organization I examined no longer exists.
A big part of the difference between the two analytical corps was that the OSINT group, except for the DI Division Chief, was not cleared for access to classified information, and for a very good reason. Most, if not all, of its analysts could not have passed the standard qualification test to receive a clearance. Virtually all had friends or relatives they were still in contact with overseas. On the plus side, they also differed in that virtually all were native speakers of Chinese, Korean, Arabic, Persian, or Russian, with the majority having received a PhD degree from a top US institution. Also, they were not focused primarily current reporting, but research projects on our principle adversaries.
I came to know this group from a project on North Korea requested by the DNI’s manager for North Korean issues at the time, Ambassador Joe DeTrani. It came about in the aftermath of the DIA underestimating the size of the North Korean ground forces by at least a third. My assignment was to determine how much research had been done prior to our discovering the mistake. My methodology consisted of reading every report in the previous ten years produced by anyone that referenced North Korea’s ground forces. My first discovery was I couldn’t find a single true research project. Several reports were described as research, but upon closer examination they all proved to be simple summaries of current intelligence reports. With this mind, I pursued on my own a similar exercise on WMD research papers produced from the Gulf War and the time of the Iraq WMD fiasco only to find that like North Korea that in this case there had been no all-source in-depth research focused on the question of Iraq’s WMD program during this critical period.
Having examined all reports published on North Korea I recognized that the OSINT analysts also often provided more insightful current reporting than the DI was producing. When I sought to dig a bit deeper I discovered that the DI analysts were jealous of the OSINT analysts, and rarely took their views into account in preparing their own reports. I found that this could be explained in large part by many all-source analysts’ preference for classified information, but in a number of cases it would also have required the DI to admit that several of its understanding of North Korean military behavior were likely incorrect. An understandable reaction maybe, but in my own experience as an analyst I found that some of my most important discovers were when I realized that my hypothesis was wrong, and I needed to rethink the puzzle I was attempting to solve.
Thinking back over my time as an analyst at DIA and later CIA, the biggest problem I encountered with using OSINT was the difficulty in asking the main collection organization, most often FBIS, a specific question or for it to put a special emphasis on a particular topic or look back at old data. To do this, I had to go through a large an complicated separate bureaucracy that most often proved to be not worth the effort. Equally dauting was the requirement of sorting through by hand myself all the mentions of a particular word or issue I was investigating. My solution in one particular case was to adopt the WWII punch card system where I could record every mention made of China’s militia, punch out each specific variable(s) mentioned and make a note on the card of my assessment of its relevance and importance to my research topic. Since no one is doing in-depth research anymore I expect few run into this problem today.
While I applaud CIA taking on a greater OSINT responsibility, I think we should broaden our I thinking. I would like to see a brand-new collection agency formed comparable to NSA or NGA exclusively for the purpose of collecting open-source information similar to what FBIS did, but also include information obtained from the internet’s social media. Much like single-source agencies such as, NSA and NGA, the new agency would produce current reports, but also be able to respond to individual research requests from all-source analysts. I think this is best done by collocating a few single-source OSINT analysts to advise and assist the all-source analysts at CIA, DIA, and INR. I know that this type of colocation worked especially well when I was at INR.