FARO stands for Frontline Record of Online Harms. Technology-related harm usually reaches the public one case at a time: a school in one country, a family in another, each easy to treat as an exception. This directory puts those cases side by side to help grassroots actors see where others face the same issues they do, and to help governments and regulators see the bigger picture of the harms occurring. The directory describes what its records contain. It does not measure how often these harms occur, and a country with few records may simply have little published about it.
How to use this directory: Topics are filterable for FARO reports that have already been created based on these incidents. The harms and response patterns shown are those featured there. To see the incidents used in a country-specific report, choose that country under Country report. Pick a pattern or evidence type to see on the map which countries have matching records. Click a country on the map or in the list to see its records below. Use the filter to search for keywords or specific sources.
Shaded countries have at least one record. Choose a pattern to see where it appears.
Shading shows only that a country has at least one record in this directory that matches your selection, and every shaded country looks the same on purpose. It says nothing about how often these harms occur there. Pattern filters only cover records classified so far, and small countries are easy to miss on a map, so use the list as well.
In the list but not drawn on the map: Multiple countries, Not stated.
A female data annotator and her husband annotated data for about a year. The platform operator stopped paying wages and six months later they had still received nothing.
An annotator reported wages were cut after a project moved to Africa and pay per microtask had halved since 2022.
A 26-year-old male worker received only $12 for working on a project over three workdays on an online data annotation platform. He had expected to earn $50 for his labor but received significantly less due to the platform's compensation practices.
A 36-year-old male data annotation worker accumulated over $150 in payments for completed work on an online data annotation platform. He was suddenly booted from the platform and lost access to his account, resulting in his earned wages being permanently withheld.
A designer based in Quezon City spent 60 hours working on 12 design contests on a creative freelancing platform. After winning only one contest for $500 and paying the platform's fees, the designer netted $350, resulting in low hourly compensation and significant unpaid work.
A group of 34 freelance data annotators in the Philippines reported experiencing delayed, reduced, or cancelled payments after completing tasks on an artificial intelligence data annotation platform. The platform's operators routinely withheld or delayed earnings without explanation or recourse, leaving many of these workers earning well below the daily minimum wage.
A 23-year-old female freelance worker was paid only 30 cents for completing four hours of work on an online data annotation platform. She experienced earnings far below the local minimum wage due to the platform's payment practices.
A virtual assistant based in the Philippines spent ₱800 on a freelancing platform's bidding system and applied to 40 jobs over the course of a month. After securing only one $200 project and paying the platform's commission fees, they netted only $160, resulting in a low financial return for a month of aggressive bidding.
A graphic designer based in the Philippines experienced low earnings on a freelancing platform due to the platform's flat 20% commission fee and client demands for unlimited revisions. For a $50 logo design, the designer netted only $40 and was forced to perform extensive revisions, resulting in compensation below the minimum wage.
An adult virtual assistant based in the Philippines experienced sustained monthly financial losses due to a freelancing platform's commission structure. The freelancer lost approximately ₱20,000 every month in service fees paid to the platform.
Each record is a short summary of an incident described in a public source or collected through an interview by SHIELD or its partners. Records are reviewed by people before they appear here. Names and identifying details are removed: where a person appears, a role, age band and gender stand in for the name, and dates and places for minors are generalised. Every record links to its original source and shows the date that link was last checked. Records are classified with shared harm categories. The classification is SHIELD's own, so it can differ from how a source describes the incident.
You are welcome to read, link to and quote individual records with credit to FARO. Please do not copy, scrape or republish the dataset in bulk, or use it to train or test automated systems, without written permission. To ask for permission, use the SHIELD contact form. The original sources keep their own rights, and the link on each record leads to them.
SHIELD-APS. Frontline Record of Online Harms (FARO). DIRECTORY-URL. Accessed [date]. To cite a particular selection, use the web address shown after you set the filters.
When you rely on the facts of an incident, cite the original source it links to. When you rely on the classification, the patterns or the country report, cite FARO. Please describe the records as documented incidents and do not present counts or country coverage as how often these harms happen. A country with few records may simply have little published about it.
If a record is wrong, a link no longer works, or you are described in a record and want it changed or removed, use the SHIELD contact form and include the title of the record. We review each request and update or remove the record where the request is justified.
Know of a documented incident that belongs here? Send it through the incident form. A public source, such as a news report, a court or regulator decision, or a research publication, helps us verify it quickly. Submissions without a public source are kept anonymous and are not tied to named products or companies. Please leave out the names of children and of anyone who has not agreed to be named. Submissions are reviewed before anything is published, and sending one does not guarantee a record.
