Author
Office of the Director of National Intelligence (ODNI)
2 reading cards from 2 books · 2015–2020.
2 cards
Intelligence Community Directive 203: Analytic Standards · 2015
An assessment can be accurate in every sentence and still false if it was tailored to a policy.
Directive 203 of the US Intelligence Community sets five standards for every analytic product: it must be objective, independent of political consideration, timely, based on all available sources, and must implement the nine analytic tradecraft standards. The text published by ODNI is the one signed on 2 January 2015, with a 2022 technical amendment that mainly concerns the role of the analytic ombuds, not the standards themselves. Objectivity means more than lack of bias: analysts must be aware of their own assumptions, use techniques that reveal bias, consider contrary information, and not stay tied to earlier judgements once the facts have changed. In practice it works as a pre-publication checklist: have we weighed alternative perspectives? Will the assessment reach the decision-maker before the decision? Have we used sources we would rather ignore? Have we said how sure we are, and why? The standards fit anywhere an assessment is written for someone who decides — a ministry, a bank or a newsroom. The typical trap is not lying but tailoring: an assessment written to please an audience, an agenda or a policy. That is why the text forbids both verbs — neither distorted by advocacy nor shaped for it.
“Analytic assessments must not be distorted by, nor shaped for, advocacy of a particular audience, agenda, or policy viewpoint. Analytic judgments must not be influenced by the force of preference for a particular policy.”
Artificial Intelligence Ethics Framework for the Intelligence Community (v1.0) · 2020
The answer to ‘how do you know?’ must lead to a source a human has read, not to the model.
In 2020 the US intelligence community published six principles of AI ethics and an implementation framework written as a list of questions. The human-centred principle asks that technological guidance be tempered with human judgement; the framework asks who the accountable human is, what they must know about the model to judge its reliability, and how AI outputs are marked. ICD 505, signed on 17 January 2025, requires data traceability from AI inputs to AI-derived outputs and systems that let analysts meet ICD 203 and ICD 206, the analytic and sourcing standards. The US AI Action Plan of July 2025 also called for an AI assurance standard under ICD 505. Applied in 2026 to a product drafted with a language model, this means: every factual claim leads to a real source that a human has opened and read, not to the model's answer; quotes and figures are checked at the source, because models invent plausible references; probability and confidence remain the analyst's judgement; and whatever the machine produced is marked. The main trap is automation bias: accepting a fluent machine's output without the checks you would demand of a new colleague. The framework says it plainly: machine errors may differ from human errors.
“How might you respond to an intelligence consumer asking “How do you know this?” How will you describe the dataset(s) and tools used to make the output?”