bans specified harmful uses of AI. Check these before you buy, release or use a tool.
The original bans already apply. New prohibitions on specified intimate and child sexual abuse material apply from 2 December 2026.
Check what the system does, who it affects and whether every element of a prohibition is present. If you rely on an exception, record its conditions and supporting evidence before approval.
The original eight categories have applied since 2 February 2025. The amendment that entered into force on 27 July 2026 adds content-related prohibitions from 2 December 2026.
For each category below, record the enabled function, data, affected people and EU market or use connection. Check the detailed legal test before concluding that a category applies.
Sources (3)
Provides the Article 5 text and summary used to list the prohibited-practice categories and their stated exceptions.
The following AI practices shall be prohibited
Official Journal source for Regulation (EU) 2024/1689, including Article 5 and the wider AI Act structure.
laying down harmonised rules on artificial intelligence
Supports the new prohibited-content category, separate provider and deployer tests, safeguards standard, scope limits, 27 July 2026 entry into force, and 2 December 2026 application date.
For each claimed exception, cite the precise provision and attach the facts that satisfy it. A general risk assessment or acceptable-use policy cannot establish an exception by itself.
Live remote biometric identification for law enforcement requires strict necessity, a permitted objective, national-law permission, a fundamental-rights impact assessment, EU registration, authorisation, notification and applicable reporting. These conditions apply to use in publicly accessible spaces. Limit the use to the necessary people, place and time.
Permitted objectives include targeted searches for specified victims or missing people; preventing specified substantial threats to life, physical safety or terrorism; and locating or identifying suspects for the (1)(h)(iii) purposes and Annex II offences. The offence must carry a national maximum custodial sentence or detention order of at least four years.
In a duly justified urgent case, request authorisation without undue delay and within 24 hours. Refusal requires immediate cessation and deletion of data, results and outputs. An adverse legal decision cannot rest solely on the identification output.
The new content rules distinguish provider capability and safeguards from deployer purpose. For providers, assess intended output, foreseeable misuse and circumvention, reproducibility, and whether reasonable state-of-the-art safeguards demonstrably prevent or sufficiently reduce the output. Accidental output during lawful use does not meet the same test as deliberate prohibited deployer use.
Non-consensual intimate material concerns realistic intimate or sexually explicit depictions of an identifiable person without freely given, specific, informed, unambiguous and explicit consent to generation or manipulation. Recital 12 describes exceptional medical diagnosis or treatment by professionals when a person cannot consent, such as an emergency, subject to rights, data-protection and medical law.
(1a) limits the foreseeable-output test to outcomes reproducible without significant technical modification, combined with inadequate safeguards to prevent the outcome and correct misuse. Article 5(1b) excludes manipulation that neither increases exposure of intimate parts nor changes the nature of depicted sexual activity. The child sexual abuse material provision retains the national-law "without right" defence.
Sources (2)
Supports the Article 5 exceptions and conditions for criminal-risk support tools, medical or safety emotion inference, biometric dataset handling, and law-enforcement biometric identification.
unless and in so far as such use is strictly necessary
Supports the separate provider and deployer tests, safeguard examples, consent and medical conditions, and exclusions from the new prohibited-content category.
A reviewer should be able to follow the system purpose, user journey, data sources, affected people and conclusion from the record.
Explain why a prohibition does not apply as well as why it does. For example, state why scoring is outside the social-scoring test or why behavioural influence does not meet the harmful-manipulation test.
Sources (2)
Supports the evidence fields by naming the Article 5 facts that determine whether a prohibited practice or exception is present.
objective evidence or clear indications
Explains the AI Act risk-based framework and identifies the original eight prohibited practices as the unacceptable-risk category before Regulation (EU) 2026/1744 added another category.
The AI Act prohibits eight practices
Use these questions for new features, purchases, data sources and biometric or profiling functions. Pause approval when a possible match remains unresolved or rests only on an unverified supplier claim.
Repeat the check when the same model enters a different workflow. Workplace, education, public-space, law-enforcement and vulnerability facts can change the result.
Sources (1)
Supports the screening questions by providing the specific Article 5 prohibited-practice wording.
publicly accessible spaces for the purposes of law enforcement
Save a dated decision for every category. If a possible prohibition remains, pause approval until the responsible legal, product, data and operating owners resolve it.
The resolution may require removing a function, restricting its use or proving that an exception applies and will be maintained in operation.
No. It bans specific uses, including specified sensitive categorisation and live remote law-enforcement identification in publicly accessible spaces unless strict conditions are met. Other biometric uses still need high-risk, transparency and data-protection checks.
Identify the system and use, the decision for each prohibition, the data and impact evidence, the applicable date and the proof for any exception.
Sources (2)
Supports the checklist structure by providing the Article 5 categories, conditions, authorisation requirements, and reporting elements.
No decision that produces an adverse legal effect
Confirms that prohibited practices sit in the unacceptable-risk tier of the Commission's AI Act explanation.
All AI systems considered a clear threat
Sorena can help you organise the facts, legal questions and evidence needed before approval.
"The following AI practices shall be prohibited"
"unless and in so far as such use is strictly necessary"
"objective evidence or clear indications"
"publicly accessible spaces for the purposes of law enforcement"
"No decision that produces an adverse legal effect"
"The AI Act prohibits eight practices"
"All AI systems considered a clear threat"
"laying down harmonised rules on artificial intelligence"
Clear answers on AI Act scope, risk categories, provider and user duties, general-purpose models, disclosures, rights, assessments and deadlines.
Find the AI Act deadline for your system or model. Check high-risk dates, transparency duties, older-model transitions and the evidence to prepare.
Check who needs an Article 27 fundamental rights impact assessment, what it must contain and how DPIA reuse, notification and registration work.
Collect model documentation, integration information, copyright records and training summaries. Add testing, incident and security records for systemic risk.
Set up high-risk AI monitoring, identify serious incidents and check reporting deadlines, recipients, deployer escalation and corrective action.
Check the AI Act fine tiers, smaller-organisation caps, GPAI provider fines and the facts authorities consider when deciding a penalty.
Check serious-incident harm, preserve evidence and choose the reporting deadline and recipient for high-risk AI systems and systemic-risk GPAI models.
See which AI Act rules apply now and what comes next. Check high-risk dates, older-model transitions and the difference between legal and standards dates.
Compare AI Act legal duties with ISO/IEC 42001 management-system requirements. See which records can be reused and what certification does and does not cover.
Compare AI Act legal duties with the voluntary NIST AI RMF. Map risk records to specific requirements and identify legal steps that need separate evidence.
Use five checks to establish AI Act scope, exclusions, your role, risk categories and evidence for the decision.
Choose between provider-run internal control, notified-body assessment and a product-law procedure before release.
Classify an AI tool or model before purchase or release. Check EU scope, prohibited uses, high-risk rules, transparency and GPAI duties.
Check each AI tool, decide which duties apply, and give every action an owner and evidence link.
Some high-risk systems use a provider-run assessment. Others need an authorised independent assessor or a product-law assessment.
Check whether the AI Act applies to your system or model, identify your provider or deployer role, and keep evidence for the decision.
Check GPAI provider duties for documentation, integrator information, copyright and training summaries, plus extra requirements for systemic-risk models.
High-risk providers must meet system requirements before supply. Deployers must use and monitor the system under the provider's instructions.
Check the system's risk controls, data, documentation, logs, instructions, human oversight and testing before release.
Find high-risk AI examples in recruitment, banking, education, healthcare, public services and products. Check Article 6 conditions and exceptions.
Build a compliance plan for each AI tool: your role, the duties that apply, the evidence needed and the person responsible.
Use EN 18286:2026 to organise provider quality-management procedures and keep evidence that they work for each high-risk system.
Build the technical file and provider records for a high-risk AI system, with owners and evidence links for each field.
Check Article 50 chatbot notices, synthetic-output marking, biometric notices, deepfake disclosures and public-interest text exceptions.
Check provider, deployer, importer and distributor roles, when changes create provider duties, and how model and system responsibilities differ.
Find the assessment route for high-risk AI, when a notified body is required, and which declaration, CE-marking and registration records you need.
Check whether software counts as AI, how models differ from systems, when embedded AI is high-risk, and which scope and role evidence to keep.
Understand high-risk AI monitoring under Article 72: the provider plan, deployer feedback, logs, corrective action, incidents and change reviews.
See which industry AI uses fall under Annex III, when Article 6(3) can apply, and what providers must document and register.
See what Article 11 and Annex IV require in high-risk AI technical documentation, from system identity and testing to monitoring and retention.
Check general-purpose model duties, systemic-risk classification, open-source conditions, copyright, training summaries, incidents and application dates.
Check AI interaction notices, technical content marking, biometric notices, deepfake and text disclosures, exceptions, accessibility and deadlines.
Check serious-incident outcomes, reporting roles, 15-, 10- and two-day deadlines, authority recipients, corrective action and the separate model route.
Check who needs an Article 27 fundamental rights impact assessment, what it must cover, and how privacy assessments, notification and registration fit.