The Federal Bureau of Investigation's Foreign Language Program -- Translation of Counterterrorism and Counterintelligence Foreign Language Material
Report No. 04-25
July 2004
Office of the Inspector General
Critical to the Federal Bureau of Investigation's (FBI) success in protecting national security is its ability to prioritize, translate, and understand in a timely fashion the information to which it has access. In this regard, the Foreign Language Program's support to the FBI's law enforcement function is substantial. In addition to supporting the FBI's two highest priorities of counterterrorism and counterintelligence, increasing demands have been placed on the Foreign Language Program to support the FBI's criminal and cyber-crimes programs, international training, international deployments, and interpreting/interviewing assignments. Prior reviews of the FBI's Foreign Language Program revealed severe shortages of linguists that resulted in the accumulation of thousands of hours of audio and videotapes and thousands of pages of text going unreviewed or untranslated. As the FBI continues to focus its priorities on counterterrorism and counterintelligence, it must rely heavily on linguistic capabilities for interview support and surveillance activities.1 The FBI's linguists play a critical role in developing effective intelligence and counterterrorism information. Linguists are the first line of analysis for information collected in a language other than English. Linguists must use their judgment in filtering the information to ensure that information of potential intelligence value is passed along to agents or analysts. Linguists must sort through the thousands of hours and pages of intercepted telephone conversations and documents to identify pertinent foreign intelligence information. Information of intelligence value is often subtle, because the parties to the conversation may suspect they are being monitored. For example, linguists must be able to recognize coded words or the implications of a conversation when the parties refer to issues cryptically. This requires high standards of language proficiency and cultural knowledge. The Foreign Language Program presents significant management challenges for the FBI, including prioritizing workload and balancing limited resources. As a result, the Office of the Inspector General initiated this audit to assess the FBI's ability to translate critical foreign language material and success at meeting linguist hiring goals and examine whether the FBI's procedures ensure the appropriate prioritization of work, accurate and timely translations of pertinent information, and adequate pre and post-hire security screening of linguists. With regard to our prioritization tests, we focused on the FBI's counterterrorism Foreign Intelligence Surveillance Act (FISA) audio material.2 Foreign Language Program Development According to the FBI, the Foreign Language Program has experienced dramatic changes over the last several years.3 In January 1999, the then Language Services Unit was elevated in organizational status to become the Language Services Section. Since then, foreign language program operations have evolved from a highly decentralized environment to one that is now highly centralized at FBI Headquarters. The Language Services Section now centrally manages a program consisting of over 1,200 linguists and a program management staff of 65 employees. The Section's scope of responsibility has significantly and substantively expanded. For example:
FBI Linguistic Capabilities The FBI's approximately 1,200 linguists are stationed across the United States in 52 field offices and Headquarters. They are connected via secure communications networks that allow a linguist in one FBI office to work on projects for any other office. As of April 2004, the 1,200 linguists consisted of about 800 contract linguists and 400 language specialists.4 Functionally, there is little difference between contract linguists and language specialists. Language specialists are employees of the FBI, while contract linguists are contractors. In addition, unlike language specialists, contract linguists are not under an obligation to be available for work. According to the FBI, contract linguists assigned to counterterrorism and counterintelligence matters work an average of 29 hours per week. FBI officials told us that the lack of contract linguist full-time availability can disrupt the normal course of investigations and an office's ability to monitor FISA lines on a near-live basis.5 Since September 11, 2001, the FBI's Foreign Language Program has experienced a large influx of funding and linguists. Language program funding increased from $21.5 million in fiscal year (FY) 2001 to nearly $70 million in FY 2004; the number of linguists has grown from 883 in 2001 to 1,214 as of April 2004.6 According to the FBI, its linguist growth occurred at the maximum rate that the FBI's congressionally earmarked funding would allow. As shown in the following chart, the FBI has increased its linguist capabilities in languages associated with counterterrorism and counterintelligence activities. Foreign Language Program Workload and Backlog Statistics The information on the translation workload and backlog presented in this report is based on the FBI's data for the foreign language program. Each FBI field office and the Language Services Translation Center is required to submit a survey quarterly to the Language Services Section for each language processed at that office. The survey requires the reporting of the volume of audio/video and text collected, forwarded to another office, received from another office, and reviewed. The quarterly survey does not require the reporting of backlog, which the FBI defines as any unreviewed material. Field offices manually compile the survey data from a number of sources. Results are recorded either manually or electronically on a WordPerfect form and mailed or e-mailed to the Language Services Section where the data from each office is entered into an Excel spreadsheet. Supervisory Language Specialists at the offices we audited with high volumes of translation work told us that it takes about a week to compile all the statistics and prepare the reports. As recognized by Foreign Language Program management, this process is inefficient and imprecise, and the use of these data is limited in that it may not accurately reflect the amount of unreviewed material. Therefore, our discussions in this report of the foreign language program's workload and backlog are presented with this caveat. However, recognizing the limitations of the survey data, we present these statistics because they are the only data available at this time. Further, these data are the same data the Language Services Section uses for budget preparation, workforce planning, and measuring success toward its strategic objective of 100 percent coverage of FBI-collected intelligence. Because of the inefficiency and imprecision of its workload survey process, the Language Services Section is working with the FBI's Information Technology Division on the development of a nationally integrated statistical collection and reporting system known as the Electronic Surveillance Data Management System (EDMS). According to the FBI, EDMS ultimately will be the presentation/reviewing system for all FISA digital data collected and will be able to monitor all information for potential backlog and duration of backlog. The system will also provide statistical reports regarding backlog. The phased rollout of EDMS is planned to begin in FY 2005. The FBI's Ability to Translate All Critical Foreign Language Material After the terrorist attacks of September 11, 2001, the FBI Director stated that the FBI needed to change from an agency primarily focused on investigating crime to one whose primary focus is the prevention of future terrorist attacks. The FBI's change in focus shifted agent investigative resources to counterterrorism and counterintelligence programs heavily dependent upon electronic surveillance collection.7 The FBI's electronic surveillance collection in languages primarily related to counterterrorism activities (i.e., Arabic, Farsi, Urdu, and Pashto) has increased by 45 percent, when comparing total collection in FY 2003 to total collection in FY 2001. Text collection in these languages has increased 566 percent, increasing from 2 percent of total collection in FY 2001, when combining audio and text collection, to 10 percent in FY 2003. Translation growth rates in these languages are expected to trend upward by at least 15 percent annually. Further, the FBI expects translation growth in languages primarily related to counterintelligence activities, [CLASSIFIED INFORMATION REDACTED], to trend upward by at least 10 percent annually. The following table depicts the increase in material requiring translation since FY 2001.
The FBI cannot translate all the foreign language counterterrorism and counterintelligence material it collects. The FBI's collection of material requiring translation has continued to outpace its translation capabilities. In fact, despite the infusion of more than 620 additional linguists since September 11, 2001, the FBI reported that nearly 24 percent of ongoing FISA counterintelligence and counterterrorism intercepts are not being monitored.11 With respect to unreviewed material, the FBI currently only tracks backlog for counterterrorism FISA cases, and the FBI only does so by case, not by language. Using available foreign language program data, we compiled the following statistics which show that since September 11, 2001, more than 89,000 hours of [CLASSIFIED INFORMATION REDACTED] audio and 30,000 hours of audio in other counterterrorism languages have not been reviewed. Additionally, over 370,000 hours of audio in languages associated with counterintelligence activities have not been reviewed, including over 116,000 hours in [CLASSIFIED INFORMATION REDACTED] and 88,000 hours in [CLASSIFIED INFORMATION REDACTED]. While the FBI also has audio material supporting criminal cases that has not been reviewed, we examined and compiled the backlog statistics for audio collected for counterterrorism and counterintelligence cases - the FBI's two highest priorities as established by the FBI Director after September 11, 2001. In addition, we focused on the unreviewed counterterrorism and counterintelligence audio material instead of text material, because text collection, although it has increased since FY 2001, only represented a small percent of the FBI's foreign language workload in the high volume counterterrorism and counterintelligence languages - 7 percent and 4 percent, respectively, in the first quarter of FY 2004. However, we recognize that translation of text material also may be of critical importance and a high priority for the FBI. |
LANGUAGE12 |
Accrued Unreviewed Audio FY 2002 through 1st Quarter FY 2004 (Hours) |
Total Audio Collected FY 2002 through 1st Quarter FY 2004 (Hours) |
Percent Unreviewed of Total Collected |
---|---|---|---|
All Languages Translated For Counterterrorism and Counterintelligence Cases |
499,611 | 1,675,268 | 30% |
Counterterrorism Language Totals |
123,638 | 617,331 | 20% |
[CLASSIFIED INFORMATION REDACTED] | |||
Counterintelligence Language Totals |
370,840 | 974,694 | 38% |
[CLASSIFIED INFORMATION REDACTED] | |||
Other Languages Translated for Counterterrorism and Counterintelligence Cases |
5,133 | 83,243 | 6% |
[CLASSIFIED INFORMATION REDACTED] |
Source: OIG developed from FBI Language Services Section workload data. Data are unaudited by the OIG. |
Note: Totals include all languages, but only individual languages with more than 1,000 hours of unreviewed audio are listed. |
We provided the backlog statistics we compiled to the FBI. According to the FBI, the Foreign Language Program is unable to review 100 percent of all FISA collection due to personnel and infrastructure shortages. The FBI stated that as additional resources become available, it fully intends to review unaddressed workload generated by many investigations should circumstances warrant. The FBI further stated that over time the intelligence value of unreviewed materials may fully degrade for myriad reasons, particularly in the counterintelligence arena, and that each FBI case is unique and judgments about the review of unreviewed materials rests with case agents and field office and operational division management officials. The FBI also stated that when these officials conclude that there may be investigative value in reviewing archived materials and linguists are available, these materials can be reviewed. With regard to the reliability of the backlog statistics, the FBI did not take exception to our use of the Foreign Language Program's workload data to compute the approximate amount of unaddressed audio supporting counterterrorism and counterintelligence investigations. However, the FBI stated that given the imprecision of the Foreign Language Program's workload reporting process, the data cannot be reliably used to determine the exact amount of unreviewed material. The purpose of our presentation is to provide the best available FBI data on the extent of unreviewed audio by language, field office, and fiscal year. As mentioned previously, we used the same data that the Language Services Section uses for budget preparation, workforce planning, and measuring success toward its strategic objective of 100 percent coverage of FBI-collected intelligence. In our judgment, the backlog statistics have value, even though we agree that the imprecision of the FBI's workload reporting process may result in statistics that may not be completely accurate in describing the exact amount of unreviewed material. In presenting the statistics, it is not our intent to imply that all previously unreviewed material should now be reviewed. In addition, we recognize that the unreviewed audio may not necessarily represent critical intelligence information or even material that requires foreign language translation.13 As described below, we attributed the FBI's backlog of unreviewed material to its difficulties in hiring sufficient numbers of linguists and limitations in the FBI's translation information technology systems. Difficulties in Hiring Linguists At a news conference with the Attorney General on September 17, 2001, the FBI Director announced the FBI's critical need for additional Arabic, Farsi, Pashto, and Mandarin contract linguists. According to the FBI, more than 20,000 applications spanning the foreign language spectrum were received by its on-line employment application collection system within the next 30 days. The FBI estimates that the contract linguist vetting process eliminates over 90 percent of the applicants processed for hiring.14 The ratio of contract linguist applicants selected for processing to hires is as follows: FY 2002 - one hire for every 14 applicants, FY 2003 - one hire for every 13 applicants, first half of FY 2004 - one hire for every 19 applicants. For those applicants who pass the vetting process and are hired, the applicant processing cycle is about 13 months.15 As previously mentioned, from September 11, 2001, through April 1, 2004, the FBI hired 626 linguists - 548 contract linguists and 78 language specialists. Nonetheless, the FBI has difficulty hiring linguists because it must compete with other Intelligence Community agencies as well as private firms. Despite the fact that the government has employed thousands of linguists, these numbers are not sufficient to keep abreast of the changing requirements of the Community's intelligence mission. Because Intelligence Community agencies are responding to similar threats, linguists in the same languages are in high demand at each agency. System Limitations In addition to hiring difficulties, we also found that system limitations contributed to the FBI's backlog of unreviewed material. The FBI's digital collection systems have limited storage capacity. Because of this, audio sessions resident on a system are sometimes deleted through an automatic file deletion procedure to make room for incoming audio sessions. Although sessions are automatically deleted in a set order, we found that sessions which are unreviewed are sometimes included in those that are deleted, especially in offices with a high volume of audio to review. Because these sessions have not been reviewed, they will continue to be reflected in backlog statistics. The FBI has been aware of problems regarding audio sessions being automatically deleted and the inability of the digital collection systems to identify or quantify the volume of deleted audio. Yet, necessary system controls have not been established to prevent critical audio material from being automatically deleted, such as protecting sessions of the highest priority on digital collection systems' active on-line storage until linguists review them. The results of our tests showed that three of eight offices tested had Al Qaeda sessions that potentially were deleted by the system before linguists had reviewed them. Although sessions deleted from an office's digital collection system can be retrieved and reimported onto the system from the copy archived on magnetic optical disks, FBI systems and foreign language officials told us that, absent a case agent indicating that there should have been more audio sessions intercepted during a certain time frame, systems and foreign language staff in an office would not know to request retrieval of any archived audio sessions. Further, archived audio sessions do not include session status information such as "Reviewed" or "Unreviewed." Thus, once audio sessions have been deleted from the active on-line storage of the digital collection systems, it is impossible for the FBI to identify from digital collection system data those audio sessions that have been reviewed and those that have not. Absent that information, sessions retrieved from archive must be reviewed by linguists even though many of these sessions may have been reviewed previously. In response to the growing demand for quality translation services, the FBI built a secure communications network, known as Arachnet, to distribute FISA audio among FBI field offices. This technology allows the FBI to more effectively utilize its linguist pool. However, Arachnet did not always successfully forward work from one office to another, adding to the backlog and causing delays in the translation of potentially critical information. For example, in one field office, systems staff were not monitoring the autoforwarding of audio sessions. As a result, unreviewed counterterrorism FISA audio sessions that were programmed to be autoforwarded to another office for translation were still on the office's system when the sessions should have been immediately forwarded upon intercept. Some of the unreviewed audio was more than two weeks old when we discovered this problem. In addition, the Language Services Translation Center at FBI Headquarters, which receives translation work from field offices, does not always receive intercepted Al Qaeda FISA audio within 12 hours. The FBI Director expects counterterrorism FISA audio to be reviewed within 24 hours of interception, and FBI policy requires Al Qaeda FISA audio to be reviewed within 12 hours of interception.16 As shown in the following table, during April 2004, 36 percent of intercepted Al Qaeda FISA audio sessions that were forwarded to the Language Services Translation Center were not received within 12 hours. |
Language |
12 Hours or Less |
Percent Received in 12 Hours or Less |
1 Day |
2 Days |
3 Days |
4 Days |
5 Days |
6 Days |
7 Days |
8 Days |
9 Days |
10 Days |
11 Days |
12 Days |
13 Days |
14 Days |
15 Days |
16 to 30 Days |
31 to 60 Days |
>60 Days |
Total Audio Sessions |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
Total Counterterrorism Sessions All Cases |
902 | 59% | 39 | 61 | 26 | 59 | 40 | 32 | 38 | 32 | 17 | 17 | 28 | 7 | 14 | 11 | 18 | 99 | 73 | 6 | 1,519 |
Total Sessions - Al Qaeda Cases |
887 | 64% | 20 | 46 | 19 | 40 | 37 | 26 | 28 | 25 | 17 | 15 | 23 | 7 | 14 | 11 | 18 | 99 | 46 | 4 | 1,382 |
[CLASSIFIED INFORMATION REDACTED] |
588 | 99% | 3 | 1 | 1 | 593 | |||||||||||||||
16 | 15% | 5 | 21 | 7 | 8 | 15 | 9 | 11 | 6 | 2 | 2 | 2 | 2 | 106 | |||||||
0% | 1 | 1 | 41 | 4 | 47 | ||||||||||||||||
283 | 44% | 15 | 24 | 11 | 29 | 21 | 17 | 17 | 19 | 15 | 15 | 21 | 7 | 14 | 11 | 18 | 97 | 2 | 636 | ||
Total Sessions - Non-A Qaeda Cases |
15 | 11% | 19 | 15 | 7 | 19 | 3 | 6 | 10 | 7 | 2 | 5 | 27 | 2 | 137 | ||||||
[CLASSIFIED INFORMATION REDACTED] |
15 | 0% | 1 | 7 | 19 | 3 | 6 | 10 | 7 | 4 | 27 | 2 | 86 | ||||||||
30% | 17 | 15 | 2 | 1 | 50 | ||||||||||||||||
0% | 1 | 1 |
Source: OIG developed from FBI digital collection system data. |
The FBI currently is pursuing initiatives designed to integrate advanced technologies into its translation systems architecture. Although these efforts are promising, more needs to be done. For example, at the Washington, D.C., field office, system maintenance contractors have taken steps to reduce the volume of calls that linguists must review by removing certain unintelligible sessions. [CLASSIFIED INFORMATION REDACTED]. This technique reduced the volume of calls to be reviewed by linguists by almost 30,000 calls lasting over 1,800 hours during the month of April 2004 alone. We recommend that this best practice and others developed by individual offices should be shared and adopted throughout the FBI. Prioritization of Workload Because the FBI does not have the ability to translate all of its foreign language material, the manner in which the FBI prioritizes its translation workload is critically important. According to the Language Services Section, it coordinates the threat-based priority of FISA coverage with operational division officials to ensure that it directs foreign language program resources toward those investigations considered to be the highest priority. Nonetheless, we found that the FBI is not providing sufficient operational information to the Language Services Section to enable it to effectively prioritize the FBI's translation workload. For example, as of June 2004, the Counterterrorism Division had not provided the Section with information identifying the relative priority of all its cases - FISA and non FISA - requiring foreign language support. Further, the FBI had not provided the Language Services Section with a complete list of Al Qaeda FISA case numbers so that it could ensure near-live monitoring of those lines. We found that the foreign language program has not prioritized its workload nationwide to ensure a zero backlog in the FBI's highest priority cases - counterterrorism cases and, in particular, Al Qaeda cases. [CLASSIFIED INFORMATION REDACTED]. Our tests found Al Qaeda material awaiting translation that should have already been reviewed based on the timeliness requirement specified in FBI policy. However, the backlogs in the offices tested were not caused by linguists in those offices translating lower priority material. Nonetheless, the existence of Al Qaeda backlog in any office indicates a need for closer monitoring by the Language Services Section so that that this work can be redistributed to other offices, if needed, to ensure its timely translation.17 Quality Control Program Inaccurate translations can have dire consequences to the FBI's intelligence gathering and investigative efforts. To better manage the risk of material being mistranslated, the FBI instituted a national Quality Control Program in January 2003. FBI policy states that all linguists with over one year of experience with the FBI must undergo an annual work review. In addition, all new linguists with less than one year of experience with the FBI are to have all of their work reviewed for the first three months, with random quarterly reviews during the remaining nine months and annual reviews thereafter. Our review of the Quality Control Program at four FBI offices found that the new-hire and annual review requirements generally were not met in FY 2003.18 The supervisors at the four sites we audited stated that they supported the quality control concept; however, operational needs often took precedence over pulling linguists off-line to complete the required quality control reviews. Further, FBI management acknowledged that they did not expect the policy to be fully implemented during FY 2003 due to the shortage of linguist resources and operational demands. Therefore, we focused our audit tests on ensuring that new-hire reviews were being conducted as required and determining whether linguists on board for more than a year had been reviewed at least once through April 2004. We found that none of the four offices reviewed 100 percent of the translations completed by new hires during their initial three months, and the offices generally were not performing quarterly reviews as required by FBI policy. For linguists onboard for more than a year, none of the four sites had reviewed all of their linguists at least once (as of April 30, 2004), as shown in the following table. |
Office | Total Active Onboard Linguists | Number of Completed Reviews19 | Percent Completed | Number of Not Satisfactory Reviews | Percentage of Not Satisfactory Reviews |
---|---|---|---|---|---|
Language Services Translation Center20 | 32 | 5 | 16% | 0 | 0% |
New York | 105 | 71 | 68% | 5 | 7% |
San Francisco | 29 | 12 | 41% | 2 | 17% |
Washington, D.C. | 92 | 63 | 68% | 10 | 16% |
Source: OIG developed from FBI sample site data. |
The FBI official responsible for the Quality Control Program acknowledged that implementation of the quality control policy faced two major problems - the Language Services Section's inability to monitor and track nationwide adherence to the policy because of lack of personnel and the field's inability to comply with the policy because managers and supervisors must use the same linguists already overburdened with operational assignments to carry out the quality control reviews. The official said that full operational coverage and the accuracy of translations are of utmost importance and one priority cannot be sacrificed for the other. The official also said that if resources become available, the solution is to establish a permanent quality control staff, independent of operational needs, whose primary duty would be to perform random quality control assessments of work produced by the linguist workforce. In addition to lack of ongoing oversight, we believe the Quality Control Program requires strengthening in areas such as reviewing audio sessions deemed not pertinent, specifying certification and review requirements for linguists that translate multiple languages, identifying responsibility for reviewing new hires on temporary duty status, providing sufficient guidance regarding management actions that should be taken when quality control review results are not satisfactory, and maintaining a tracking system of all quality control reviews. Post-Adjudication Risk Management The FBI has a difficult challenge in balancing the need for qualified linguists with often-competing personnel security concerns. To address this issue, the FBI instituted a post-adjudication risk management (PARM) program in October 2002 that mandates periodic personnel security interviews and polygraph examinations for contract linguists where the inherent risk to the United States is higher than usual. As of May 2004, 193 of approximately 350 PARM reviews had been completed, 108 were partially completed, and 48 had not been started. Of the 193 PARM reviews completed, 12 contract linguists had been suspended due to deceptive or inconclusive polygraphs. Two of these linguists were reinstated after a second polygraph or upon further investigation. The investigations are still pending for the ten remaining linguists. Recommendations This report contains 18 recommendations to help improve the FBI's Foreign Language Program. Among these recommendations are that the FBI should:
FBI officials were receptive to the recommendations we offered during the audit and in many instances took corrective actions while we were on site. Footnotes
* This is an unclassified executive summary of the Office of the Inspector General's (OIG) 157 page audit report, which was classified "Secret" by the Federal Bureau of Investigation (FBI). To create an unclassified executive summary for public release, the OIG redacted (whited out) the portions of the full executive summary that were classified by the FBI. The redacted portions are noted, and included references to specific languages and FBI field offices, and the identification of certain capabilities and vulnerabilities. The full report was provided previously to the Department of Justice, the FBI, and several Congressional committees. REDACTED AND UNCLASSIFIED |