The report presents the findings of a CETaS research project commissioned by the Joint Intelligence Organisation (JIO) and GCHQ, on the topic of artificial intelligence (AI) and strategic decision-making.
The report assesses how AI-enriched intelligence should be communicated to strategic decision-makers in government, to ensure the principles of analytical rigour, transparency, and reliability of intelligence reporting and assessment are upheld. The findings are based on extensive primary research across UK assessment bodies, intelligence agencies, and other government departments, conducted over a seven-month period throughout 2023-24.
‘AI-enriched intelligence' in this context refers to intelligence insights that have been derived in part or in whole from the use of machine learning analysis or generative AI systems such as large language models.
The research considered:
- Whether national security decision-makers are sufficiently equipped to assess the limitations and uncertainty inherent in assessments informed by AI-enriched intelligence.
- When and how the limitations of AI-enriched intelligence should be communicated to national security decision-makers to ensure a balance is struck between accessibility and technical detail.
- Whether further governance, guidelines, or upskilling may be required to enable national security decision-makers to make high-stakes decisions based on AI-enriched insights.
Key findings from the research:
- AI is a valuable analytical tool for all-source intelligence analysts. AI systems can process volumes of data far beyond the capacity of human analysts, identifying trends and anomalies that may otherwise go unnoticed. Choosing not to make use of AI for intelligence purposes therefore risks contravening the principle of comprehensive coverage in intelligence assessment, set out in the Professional Head of Intelligence Assessment Common Analytical Standards. Further, if key patterns and connections are missed, the failure to adopt AI tools could undermine the authority and value of all-source intelligence assessments to government.
- However, the use of AI exacerbates dimensions of uncertainty inherent in intelligence assessment and decision-making processes. The outputs of AI systems are probabilistic calculations (not certainties) and are currently prone to inaccuracies when presented with incomplete or skewed data. The opaque nature of many AI systems also makes it difficult to understand how AI-derived conclusions have been reached.
- There is a critical need for careful design, continuous monitoring, and regular adjustment of AI systems used in intelligence analysis and assessment to mitigate the risk of amplifying bias and errors.
- The intelligence function producing the assessment product remains ultimately responsible for evaluating relevant technical metrics (such as accuracy and error rates) in AI methods used for intelligence analysis and assessment, and all-source intelligence analysts must take into account any limitations and uncertainties when producing their conclusions and judgements.
- National security decision-makers currently require a high level of assurance relating to AI system performance and security to make decisions based on AI-enriched intelligence.
- In the absence of a robust assurance process for AI systems, national security decision-makers generally exhibited greater confidence in the ability of AI to identify events and occurrences than the ability of AI to determine causality. Decision-makers were more prepared to trust AI-enriched intelligence insights when they were corroborated by non-AI, interpretable intelligence sources.
- Technical knowledge of AI systems varied greatly among decision-makers. Research participants repeatedly suggested that a baseline understanding of the fundamentals of AI, current capabilities, and corresponding assurance processes, would be necessary for decision-makers to make load-bearing decisions based on AI-enriched intelligence.
This report recommends the following actions to embed best practice when communicating AI-enriched intelligence to strategic decision-makers:
- The Professional Head of Intelligence Assessment (PHIA) should develop guidance for communicating uncertainty within AI-enriched intelligence in all-source assessment. This guidance should outline standardised terminology to be used if articulating AI-related limitations and caveats to decision-makers. Guidance should also be provided on the threshold at which assessments should communicate the use of AI-enriched intelligence to decision-makers.
- A layered approach should be taken by the assessment community when presenting technical information to strategic decision-makers. Assessments in a final intelligence product presented to decision-makers should always remain interpretable to non-technical audiences. However, additional information on system performance and limitations should be available on request for those with more technical expertise.
- The UK Intelligence Assessment Academy should complete a Training Needs Analysis on behalf of the all-source assessment community to identify the requirement for training for new and existing analysts. The Academy should work with all-source assessment organisations to develop appropriate training in response to the Analysis.
- Training should be offered to national security decision-makers (and their staff) to build their trust in assessments informed by AI-enriched intelligence. Decision-makers should be given basic briefings on the fundamentals of AI and corresponding assurance processes.
- Short, optional expert briefings should be offered immediately prior to high-stakes national security decision-making sessions where AI-enriched intelligence underpins load-bearing decisions. These sessions should brief decision-makers on key technical details and limitations, and ensure they are given advanced opportunity to consider confidence ratings. These briefings should be jointly coordinated by the JIO and National Security Secretariat and should draw from cross-governmental expertise from the network of Chief Scientific Advisers and relevant Scientific Advisory Councils. Guidance on when to offer briefings should be produced, and the need for briefings should be continuously assessed; as decision-makers become more comfortable with consuming AI-enriched intelligence, the level of desired assurance may reduce, and briefings may eventually become unnecessary.
- A formal accreditation programme should be developed for AI systems used in intelligence analysis and assessment to ensure models meet minimum policy requirements of robustness, security, transparency, and a record of inherent bias and mitigation. Technical assurance for the application of a system to a specific problem should be devolved to relevant organisations, and each organisation's assurance process should be accredited. This programme will require dedicated resourcing, bringing together understanding of intelligence assessment standards and processes with technical expertise. PHIA should assist in developing principles and requirements, while technical expertise for accreditation and testing should be drawn from technical authorities in the intelligence community and across government.
To read the full report, please click here.