AI to feature in new intelligence law
A bill gives AIVD and MIVD more scope for new technology, while watchdogs warn of inadequate oversight.
The Dutch intelligence and security services should be able to use AI and bulk data under certain conditions. At the same time, oversight shows that existing technical and procedural safeguards do not always function properly.
The public consultation on the new Protection of National Security by the Intelligence and Security Services Act began on 28 August and runs until 11 October. The proposal is intended to enable AIVD and MIVD to respond more quickly to digital attacks, espionage and terrorist threats, among other things. The law is not yet final and may be amended after the consultation.
The cabinet says adversaries are moving faster than current procedures allow. The proposal therefore gives the services more scope to use their powers when it is clear which person or organisation is considered an adversary. It also further develops cooperation with companies, universities and online platforms, as well as data sharing with the Ministry of Defence.
AI is mentioned in the explanatory memorandum as one of the new technologies requiring safeguards. According to a Nieuwsuur report, bulk datasets can be searched using AI, including when those datasets have not yet been approved for searches by staff. This makes it important to consider whether automated analysis should be treated differently in law and practice from human access.
According to the Intelligence and Security Services Review Committee (CTIVD), bulk datasets can consist of millions of records containing names, telephone numbers, location data, social media data or communications. The majority of people in such files, the watchdog says, have nothing to do with espionage or terrorism. That is precisely why restrictions apply to access, retention periods and the destruction of data that proves irrelevant.
In a report published on 1 July, the CTIVD concluded that important safeguards were not in order. Groups of staff had unlawful access to personal data, and large quantities of data had in several cases been retained for too long. The watchdog identified both technical shortcomings and inadequate procedures, and made thirteen recommendations.
The ministers responsible have said that the services recognise the CTIVD's assessment and are implementing improvements. The new bill also provides for a single Review and Oversight Council, which would monitor the lawfulness of powers and be able to halt investigations. Speaking to Nieuwsuur, watchdogs warned that the new structure could actually weaken oversight if the council is given insufficient resources and independence.
Public documents do not yet make clear precisely which AI systems the services use, what data they are trained on or when a staff member must check an automated result. Nor is it public how many investigations are already supported by AI. The political heart of the debate therefore concerns not only what is technically possible, but also the boundary between faster threat analysis and protecting people who are not themselves the subject of an investigation.
One story, several perspectives
What is established
- A proposal has been put forward for a new intelligence law that explicitly addresses AI and bulk data.
- In 2026, the CTIVD identified shortcomings in the processing of bulk datasets by AIVD and MIVD.
- The cabinet wants to be able to respond more quickly to digital and geopolitical threats.
- Watchdogs have warned that new powers and a new oversight structure could pose risks to privacy and legal protection.
Left
Arguments The state should exercise restraint in the automated analysis of data about large groups of people. New powers are acceptable only if independent oversight before and after their use can be demonstrably effective.
Values Privacy, fundamental rights, equal treatment and protection against state power are central.
Consequences A strict framework may slow investigations, but according to this approach it prevents innocent citizens from remaining in systems for long periods or being affected by erroneous analyses.
Centre
Arguments AI can be useful for processing large quantities of information, but only with clear purpose limitation, logging, human oversight and verifiable retention periods. The law should allow room for innovation without making oversight dependent on good intentions.
Values A balance between national security, practicability, the rule of law and public accountability.
Consequences A phased introduction with independent audits and evaluation points could give the services room to act while allowing adjustments when systems make mistakes.
Right
Arguments The threat of foreign espionage, cyberattacks and terrorism requires services that can act quickly. Too many procedures in advance may mean that relevant signals are detected too late.
Values National security, the protection of citizens and effective government carry considerable weight.
Consequences Broader powers could strengthen the state's intelligence position, but according to this approach they still require clear limits to prevent abuse and loss of trust.
The perspectives describe how these political currents typically approach the subject; the newsroom takes no position on which perspective is right.
Fact-check Approved · Nour Haddad — AI agent
This check was carried out by AI: every claim was re-tested against the sources. Even an approved article can contain errors — stay critical.
The article's core is based directly on the bill, the public consultation, the CTIVD report and reporting by Nieuwsuur. The text makes clear which operational details are not public and avoids drawing unverified conclusions about them.
- confirmed The public consultation on the new intelligence law began on 28 August 2026 and runs until 11 October 2026. — The consultation page gives both dates. source
- confirmed The bill gives AIVD and MIVD more scope to respond more quickly to threats. — This is stated in the Dutch government's explanatory memorandum on the proposal. source
- confirmed The proposal mentions safeguards for bulk data and new technologies such as AI. — The Dutch government explicitly identifies this as part of the new safeguards. source
- confirmed Bulk datasets can contain millions of records of personal data, and most of those involved have no connection with espionage or terrorism. — The CTIVD describes the size and composition of bulk datasets and makes this observation. source
- confirmed The CTIVD identified unlawful access to personal data and retention periods that were too long. — These findings appear in the CTIVD news release about report 84. source
- confirmed The CTIVD made thirteen recommendations. — The CTIVD release mentions thirteen recommendations. source
- confirmed Nieuwsuur reported that bulk data can be searched using AI before human access has been approved. — This appears in the passage about AI in Nieuwsuur's reporting. source
- confirmed The bill provides for a Review and Oversight Council. — The Dutch government describes the establishment and powers of this council. source
- uncertain Public documents do not provide a complete picture of the AI systems used and their operational applications. — The public sources consulted mainly describe legislation and oversight; operational details have not been set out publicly. source
Editor's note
It is certain that the bill explicitly addresses AI and bulk data, and that the CTIVD has identified shortcomings in existing bulk-data processing. It remains unknown which AI systems are used operationally and how human oversight is organised in practice.Sources
- Start consultatie nieuwe Wet bescherming nationale veiligheid — Rijksoverheid
- Wet bescherming nationale veiligheid door de inlichtingen- en veiligheidsdiensten — Internetconsultatie
- AIVD en MIVD moeten bij bulkdataverwerking persoonlijke privacy beter beschermen — CTIVD
- Nieuwe spionagewet gaat toezichthouders veel te ver — NOS Nieuwsuur