Our Search Policy

How SafeShelf evaluates catalog searches and what our library's safe search principles are — published here for transparency.

General Principles

Version 1 · Effective: September 29, 2026

Our library embraces freedom of access to information as a core principle, in line with the IFLA/UNESCO Public Library Manifesto. The SafeShelf safe-search layer does not remove any item from the catalogue; it only declines to display results for searches that clearly seek methods of committing crimes against public health or public safety. Evaluation is intent-based, not topic-based: protective and educational searches such as "preventing child abuse" are free; requests for methods of committing a crime, such as "how to abuse a child", are blocked. All decisions are logged anonymously and can be appealed with a single click.

Exceptions for Researchers

Version 1 · Effective: September 29, 2026

Queries for academic, professional (health, law, security, social work) and historical research purposes should not be blocked. Textbooks, literary works and reference sources in fields such as history, literature, chemistry or medicine remain accessible. If you believe a decision is wrong, use the "I think this decision is wrong" link or contact library staff; the query can be added to the allow list after review.

Privacy and Logging

Version 1 · Effective: September 29, 2026

Searches are logged anonymously in accordance with data-protection legislation (KVKK). No IP address or identity information is stored; only the query text, the decision and a timestamp are kept. Logs are reviewed by administrators solely to improve the safe-search rules and are deleted automatically after the configured retention period.

How Does It Work?

Your query is first compared against the allowlist; educational/preventive phrases pass directly.
Spelling tricks (e.g. digit-letter substitution) are normalized and the library's pattern rules are applied.
If no rule triggers, an AI-assisted intent analysis evaluates the query using the library policy as context.
Queries found safe are searched in Open Library (openlibrary.org), the open catalogue run by the Internet Archive; for risky ones no results are listed and an information card is shown.

No filter is 100% accurate. If you believe a search was blocked in error, use the feedback link on the warning card or ask a librarian.

Safety Topics Evaluated

These topics are evaluated not as subjects, but only with respect to queries carrying an intent to cause harm.

CSAChild abuse and child sexual exploitation

Seeking abuse methods or abuse material. Prevention, warning signs and protective information are not blocked.

EXPExplosives, bombs and weapon manufacturing

Seeking methods to produce bombs, explosives or improvised weapons. Historical and literary research is not blocked.

TERTerrorism and violent extremism

Attack planning, organisation propaganda and recruitment methods.

CBRNChemical, biological, radiological and nuclear threats

Production and dispersal of poisons, biological agents or radioactive materials.

DRGDrug production and trafficking

Substance synthesis, production and sales methods. Addiction treatment and prevention information is not blocked.

SHSelf-harm and suicide methods

Method seeking; the person is directed to support resources. Prevention, psychology and literary research are not blocked.

HMNHuman trafficking and sexual exploitation

Smuggling and exploitation methods. Victim rights and anti-trafficking literature are not blocked.

CYBCyber attacks and computer crime

Unauthorised intrusion, malware creation and fraud methods. Cybersecurity education is not blocked.

PUBActs threatening public health

Methods for acts such as poisoning food or water or spreading disease.

STAThreats to state and critical infrastructure security

Sabotage of critical infrastructure; planning attacks on state institutions.