Preventive intelligence: where does warning end and unlawful surveillance begin?
European artificial intelligence rules draw an important boundary when predicting that someone will commit a crime: the assessment must not rest solely on profiling or personal characteristics. Commission guidance distinguishes this from assisting a human assessment already grounded in objective, verifiable facts connected to criminal activity. This is a European example within its legal scope, not an Iraqi rule or a prohibition on every form of risk analysis. It nevertheless raises a question extending beyond technology: when is prevention a response to a danger supported by evidence, and when does it become a way to classify people before they have done anything? Commission guidance, paragraphs 184–193.
Iraq needs early warning, as does any state confronting security risks. Waiting for an attack is not an adequate standard of protection. Yet “prevention” alone does not answer questions about authority or make all accessible information a legitimate target for collection and retention. The difference between prevention and unlawful surveillance concerns the grounds and limits of intervention, who authorises it, and what happens to information afterwards. Without these questions, institutions may expand their capacity to know while citizens lose the capacity to challenge, with no demonstrated improvement in safety.
A legitimate purpose is not sufficient by itself
In Iraq, Article 40 of the Constitution requires legal and security necessity and a judicial decision for communications surveillance or disclosure. Published constitutional text. Its practical significance is that authority does not arise from technical capability or an announced general purpose. I am not offering a judicial interpretation of every possible application, but a governance principle: the requesting body must explain why a specific purpose requires a specific intrusion, and a competent authority must be able to refuse or restrict the request and review what was done under it.
This differs from invoking “security reasons” alone. The phrase may describe a field of work, but it does not define duration, relevant people and information, or conditions for termination. If an implementing agency determines necessity, supervises itself and rules on complaints against itself, functions requiring balance become concentrated. Separating them need not obstruct urgent decisions; it requires lawful procedures that accommodate urgency without making it permanent. Proposed exceptions still require legislation and safeguards, rather than assuming that wanting an exception is sufficient evidence of its legitimacy.
It is also useful to distinguish assessing a general risk from imposing an adverse consequence on an identified person. An institution may study crime trends through aggregated data to understand service needs, but moving from a general pattern to individual suspicion requires additional grounds. Belonging to a neighbourhood, category or social network does not itself establish criminal behaviour. This concerns the quality of inference as well as rights: replacing specific evidence with a broad generalisation increases the risk that people bear the cost of a pattern that does not explain their actions.
After authorisation: the institutional life of data
In Big Brother Watch and Others v. the United Kingdom in 2021, the European Court of Human Rights emphasised safeguards throughout interception, including independent authorisation, supervision and subsequent review. It did not find that every bulk interception regime was inherently unlawful; it examined safeguards and identified deficiencies in the system before it. This European judgement is not a source of Iraqi powers. Its comparative value lies in rejecting the reduction of legality to a signature at the beginning of the process. Judgement of 25 May 2021.
Information may lawfully be obtained for one purpose while transfer to another purpose or access by unnecessary personnel creates harm. I therefore propose examining the entire information lifecycle: collection grounds, access limits, retention periods, reasons for transfer, and correction or removal when the legal need ends. These are general accountability questions, not operational details about secret collection methods. They also protect institutions against turning archives into permanent reserves for uses whose legality has never been examined.
This requires the ability to establish who used information and why, while restricting access to those whose responsibilities require it. Ethical instructions are insufficient if misuse cannot be detected or every access can be justified after the event. Contracting with a technology supplier does not transfer public responsibility to that company. The state should understand what a tool can and cannot do and which contractual rights are needed for audit and data protection. I am not claiming these conditions are present or absent in a particular Iraqi contract; they are proposed tests before approving further expansion.
Technology can amplify error as well
The UN human rights office’s 2022 report on digital privacy warned about expanding surveillance and emphasised legality, necessity, proportionality and oversight. It addressed spyware, public surveillance and encryption protections. This is an international human rights framework, not an assessment of an Iraqi system or an estimate of how many people are monitored here. Report A/HRC/51/17, particularly its recommendations. My inference is that acquiring additional capability requires testing need and harm, rather than assuming that everything collectable should be collected.
Algorithmic outputs make the issue particularly clear. A numerical result may appear more objective than an official’s opinion while still depending on human definitions, data and choices. If the data fail to represent the population, or record institutional activity more faithfully than actual harm, the system can reproduce those limitations. This is a methodological possibility to test, not a claim of demonstrated bias in a particular Iraqi system. Placing a human at the end of the process does not remove the problem if that person routinely approves the output without time or authority to examine it.
This explains the importance of the European distinction introduced above. The objection is not to calculation itself, but to imposing an individual security consequence through characteristics insufficient to establish conduct or danger. In Iraq, I would require any technological trial affecting rights to operate within a stated, bounded purpose, with its errors and costs reviewed before expansion. A supplier’s marketing claims about accuracy should not be treated as verified local performance, nor should the absence of independent evidence be remedied by saying that the details are secret.
Necessary secrecy need not prevent publication of general rules or aggregate oversight findings. Types of authorisation, oversight bodies, complaints routes and numbers of substantiated breaches can be explained without exposing an ongoing case or a source’s identity. This is not a claim that every detail should be public: disclosure can harm individuals or a lawful investigation. However, an authorised independent reviewer should have access to what is needed for verification. Otherwise, oversight becomes a request to trust the institution’s description of itself. Reviewer independence and access powers are complementary requirements, not alternatives.
Prevention that can be justified to citizens
For Iraq, I propose beginning with a legal and institutional review of purposes and powers, followed by an assessment of any new system’s effects on rights and decision quality. Competent authorities would participate within their mandates, with approval, review and complaints responsibilities defined in advance. Expansion should require evidence of a need that cannot be addressed through a less intrusive means, alongside reviewable time limits. These are analytical recommendations, not an announcement of enacted legislation or a description of a government programme. Their value is to place the burden of justifying expansion on the requesting authority, instead of expecting citizens to prove harm they may not know has occurred.
This complements the theme of internal security, intelligence and law enforcement, while remaining distinct from improving assessment accuracy alone. Information can be correct but unlawfully obtained or used; an authorised practice can also be of little value. Legality and effectiveness therefore need separate examination before their findings are combined in a decision. A system that improves one and disregards the other remains deficient, however technologically advanced or well organised on paper.
Lawful warning crosses its boundary when a possible danger becomes an open-ended licence to intrude into people’s lives, or when no independent route remains to review the grounds and continuation of that intrusion. Protecting Iraq requires sound knowledge, but also clarity about the limits of those who hold it. The strongest prevention is one whose purpose and boundaries the state can explain and whose legality it can defend, rather than one that asks society to surrender the right to question in order to feel safe.
Ali Zuweid — Researcher in strategic affairs and national security
Cite this article
Ali Zuweid. Preventive intelligence: where does warning end and unlawful surveillance begin?. 12 November 2026. https://alizuweid.com/articles/preventive-intelligence-and-lawful-surveillance/?lang=en
