AI and Civil Society: Threats and Obstacles to Deployment and Advocacy
KEY FINDINGS
- Surveillance of public and digital spaces pose severe risks to civic space. FRT deployment facilitates mass surveillance in public spaces and severely undermines anonymity, freedom of movement, and freedom of peaceful assembly. AI tools also provide law enforcement agencies and government authorities with the capabilities to collect and analyse vast amounts of data on their citizens, and those they perceive to be a threat, for instance, through AI data mining, predictive policing, and risks analysis tools. Despite the risks and biases associated with these tools, states continue to acquire them from companies looking to profit and then deploy them in human rights violations. In contexts where robust data protection and AI regulations are lacking, civil society organisations (CSOs) are even more vulnerable to the chilling effect of surveillance. As a result of increased deployment of AI in surveillance, civil society actors feel monitored and unsafe. Fearing retaliation, they may resort to self-censorship and avoid protesting or engaging in civic life and public matters.
Surveillance of public and digital spaces pose severe risks to civic space.
- Content moderation systems used by social media platforms have long been shown to impact human rights and the civic space, as they can lead to over-moderation of the legitimate right to free speech, hindering access to information and civil society’s ability to mobilise and reach audiences. This is due to a number of factors, especially: 1) embedded biases in these systems which tend to disproportionately affect Global South users; 2) while systems can categorise the type of content posted, they may still fail at determining the context in which it is posted, particularly with sensitive content, for example, health information or documentation of human rights abuses; and 3) the lack of human oversight in algorithmic content moderation makes it difficult for those affected to understand why their content was removed and then to address it.
- Under-moderation of harmful content could put the safety and wellbeing of civil society activists, journalists, and excluded groups at risk. Content recommendation systems deployed by social media platforms can exacerbate the spread of misinformation and disinformation, as well as controversial and sensationalist content such as hate speech and incitement. As a result, access to credible information and open public debate is undermined. Content recommendation systems also downrank the visibility and reach of civil society content.
- The use of AI to spread disinformation and target and discredit civil society is a serious concern. Many survey respondents said they were concerned how these tools could be used to distort narratives about their work and negatively influence public debate. Further, deepfake sexual abuse against women is contributing to their exclusion from civic participation due to safety concerns.
- Most CSOs that responded to the survey mentioned that they use AI in very limited ways, mostly to assist with research tasks, and some do not use it at all. Only a small fraction of organisations (13 percent of respondents) reported that their organisations have experience with developing their own AI solutions, such as chatbots and AI models, for data analysis or training datasets.
- Based on the survey results, a lack of financial resources and technical capacity are major obstacles to AI adoption by CSOs, followed by the lack of localised models that understand local contexts and languages.
- Nearly a third of survey respondents cited human rights and ethical concerns as one of the obstacles that prevented their organisations from effectively deploying AI or deploying it at all. Respondents were particularly concerned about the potential exacerbation of misinformation and disinformation (60 percent of all respondents) and risks to the right to privacy associated with the design and use of AI models and systems (56 percent).
- There is a lack of policies and risk mitigation measures to ensure that AI does not infringe on human rights. Of the 230 respondents who said their organisations deploy AI, only 38 percent stated that they have considered and attempted to address issues of ethics and rights in the use of AI. This can be explained by two factors. First, for those that deploy AI tools, its usage remains very limited, while only a small fraction has developed their own models. Second, CSOs lack the technical expertise and resources to address these risks.
- Measures organisations take to protect human rights with AI include protecting personal and sensitive information, conducting assessments of human rights and data audits, ensuring transparency and rights protection in automated decisions, outputs, and stakeholder engagement.
- Several obstacles prevent CSOs, particularly in the Global South, from engaging in AI advocacy and accountability efforts. First, there is a lack of financial resources and the necessary knowledge and expertise. Second, there is, in some countries,an unfavourable regulatory and political environment, as well as algorithmic opacity and secrecy from both deployers and developers of AI systems that shape AI governance and accountability.Further, CSOs face exclusion from national and international discussions and forums on AI, and for those based in the Global South, expensive travel costs, strict visa rules, predominance of the English language and lack of remote participation alternatives present additional hurdles to engagement.