Input for the Independent Expert’s report on AI and international solidarity to be presented to the 3rd Committee of the UN General Assembly in October 2024
submission prepared by Taiwan Association for Human Rights (TAHR)
TAHR
Founded in 1984, Taiwan Association for Human Rights (TAHR) is one of the most important and oldest human rights NGOs in Taiwan. TAHR has participated in many important democratic and human rights movements in Taiwan.
Website: https://www.tahr.org.tw
Contacts: info@tahr.org.tw, kuan@tahr.org.tw
2. Do you have access to procedures to communicate solidarity issues directly to AI Companies or State Institutions? What kind of procedure do you have? If not, what kind of procedure would you like to have?
2.1 The procedure should be based on human rights discussions to include vulnerable groups and civil society supports for democracy.
2.2 To make the procedure decision-influential, the opinions and inputs relating to prevention of human rights violation and mitigation of harms should be put into the decision process. The improvement should be followed up by an independent oversight body.
2.3 The procedure should exist in every stage of the AI, from planning to applications. For a partially negative example in Taiwan, the judicial body started to collect opinions from civil society “after” it faced criticism when releasing the news that the judicial body has developed AI and will deploy AI systems to help judges draft their judgment.
6. Is there transparency in the application of AI in contexts affecting you?
The transparency of the state's application of AI is not enough nor unaccountable. For example, in Taiwan there is a predictive AI being used by probation officers to evaluate the risk of recidivism, but the information of the development, deployment, and human control are not clear. Civil society does not know if there is an independent audit. There is no procedure to make sure or examine if there exist any discrimination or human rights risks.
8. Is there independent oversight of the use of AI affecting you?
There is no independent oversight of the use of AI in Taiwan. At minimum, there should be an independent oversight authority of data protection capable of regulating and giving guidelines on collecting and processing personal data during different levels of planning and deploying AI.
9. Are you subject to surveillance, disinformation, denial of benefits, preventive exclusion from engaging in cyber solidarity actions, etc.?
9.1
Plaintiffs challenged the restriction of opt-out rights and the legitimacy of mandatory secondary use of health data was put under the hat of “free rider of scientific research” and was attacked as being the “Broken Windows”. TAHR is one of the plaintiffs.
To boost the research and development of precision medicine as well as AI in the medical field, the Taiwan government provides National Health Insurance Data to third parties to do research without patients’ consent (and even restrict opt-out rights till now). In 2022 the constitutional court decided the government should set the law to respect opt-out rights and regulate secondary use of massive databases in the government.
9.2
The Taiwan government originally planned to embed “forced-consent” in the Next-Generation Sequencing (NGS) test to patients who have cancer. The NGS test was incorporated into the payment of National Health Insurance, while the government planned to force hospitals to collect patients’ consent to upload their data (including gene data) to the National Health Insurance Database to get the payment of the test. It would end up making patients be forced to agree to upload their sensitive data to broadly third party use or only patients with better economic status can afford the test. It generates medical inequality.
10. Do you have recommendations for the UN and the Independent Expert on Human Rights and International Solidarity to strengthen AI International Solidarity?
10.1 Overall
TAHR recommends the UN and the Independent Expert on Human Rights and International Solidarity to set minimum standards on different levels of developing AI, including:
-
collection and process of data
-
training and testing models
-
deployment of AI
10.2 Development and Use of AI From the Governments
Before the process of data, at minimum, the governments should assess the human rights impacts with civil society and disclose the plan.
Explicit regulation, impact assessment and rights protection mechanisms are required when the development involves sensitive data (e.g., health data) or massive amounts of data (e.g., national database).
When collecting and processing data, the governments should respect data subjects’ digital autonomy. It is necessary for governments to inform the data protection rights to data subjects, and to make efforts to raise accessibility for people to exercise the data protection rights. These rights are including but not limited to right to delete, right to access and opt-out.
When the development involves the secondary use of data, which means to process the data beyond the original purposes, the governments should seek for free, prior and informed consent (not forced consent, i.e., if a person refuses to consent, then the person should pay more for the same medical treatment. It is a forced consent.) If there are huge public interests that need data to fulfill, the governments should establish a legal framework and procedure to identify the scope of the public interests and categorize the urgency and necessity of the processing of data to alleviate data processors’ obligation of getting data subjects’ consent. In this situation, opt-out rights become the final resort that data subjects may exercise to control their data to protect privacy.
