Digital platforms still fail to protect the integrity of elections, study says
Survey analyzes the policies of social media apps, messaging apps, and artificial intelligence tools for the 2026 elections
www, internet,código binário
The main digital platforms still present significant gaps in the protection of electoral integrity in Brazil, particularly given the expansion of the use of generative artificial intelligence (AI). This is the conclusion of the study by Sala de Articulação Contra Desinformação (SAD).
The report is signed by 66 civil society organizations, laboratories, and research institutions, including Conectas, Instituto Marielle Franco, Data Privacy Brasil, and InternetLab.
The survey analyzed the policies of Facebook, Instagram, Kwai, LinkedIn, TikTok, X, and YouTube, in addition to WhatsApp, Telegram, and five generative artificial intelligence platforms: ChatGPT, Claude, Gemini, Llama, and Grok. The survey identified problems mainly in the accessibility of rules, reporting channels, and transparency on specific measures for Brazilian elections.
Read more
AI expands challenges for elections
One of the main warnings of the study is the absence of specific and easily accessible electoral policies for AI tools. Llama and Grok do not have specific and identifiable electoral policies, while information on Gemini is spread throughout multiple Google pages. ChatGPT and Claude have specific policies, but also contain transparency gaps on risk assessments, audits, and the effectiveness of safeguards.
According to the study, the widespread adoption of generative AI has changed the information landscape compared to the 2022 elections. Tools capable of producing text, images, video, audio, and conversational agents at scale lower content production costs while increasing the risks of disinformation, identity fraud, audiovisual manipulation, and the automation of political communication.
Consequently, the study highlights the need to include AI platforms in the debate on electoral regulation and to establish permanent mechanisms for governance, transparency, and accountability.
Social media has more developed, but still insufficient policies
The landscape regarding social media platforms is considered more mature, particularly concerning electoral integrity. YouTube, TikTok, and Meta maintain pages consolidating information on policies and moderation processes applicable to elections. Nevertheless, the study indicates that these rules are often scattered across various documents and that significant differences exist among platforms regarding which content should be removed, labeled, or have its reach reduced.
There are also gaps in how political violence based on gender and race is addressed. Facebook, Instagram, X, TikTok, YouTube, and LinkedIn do not recognize this phenomenon as a distinct category, instead relying on broader policies concerning hate speech, harassment, electoral integrity, and misinformation. Kwai stands out as an exception, as it has a specific policy addressing political violence against women in an electoral context, although it does not cover other dimensions, such as race.
Climate and socio-environmental misinformation, in turn, receive little attention. Facebook, Instagram, X, and LinkedIn lack specific policies on the subject. YouTube and Kwai have partial measures in place, while TikTok is the only platform analyzed that explicitly addresses the issue, including within its advertising guidelines.
WhatsApp and Telegram have differences
Among messaging apps, the study identifies a significant difference between WhatsApp and Telegram. WhatsApp has a specific policy for elections and employs mechanisms such as limits on message forwarding, identification of frequently forwarded content, restrictions on mass messaging, and the ability to block and report accounts.
Conversely, Telegram lacks a specific election policy, relying primarily on general anti-spam rules. The study also highlights a gap common to both platforms: no specific provisions were found regarding the use of chatbots, avatars, and synthetic content to simulate communication between campaigns and real people, as addressed by electoral legislation.
Lack of transparency is a cross-sectional challenge
The overall assessment, conducted in July 2026, considered three aspects: the accessibility of policies, the availability of reporting channels, and transparency regarding Brazilian elections. None of the platforms analyzed received the maximum score across all three criteria. Meta and YouTube achieved the highest average scores, at 0.8, while Telegram and Grok received an average of zero.
The study indicates that many platforms do not clearly state the publication or update dates of their rules, while reporting procedures may require navigating through multiple pages. This difficulty also limits the ability of users, researchers, authorities, and civil society organizations to understand the rules and monitor their enforcement.
Recommendations
The recommendations presented include the creation of specific, accessible webpages regarding electoral policies; the strengthening of channels for reporting and challenging content; the adoption of standardized mechanisms for identifying and tracing AI-generated content; and the conduct of periodic impact assessments, independent audits, and security tests.
The document also recommends that platforms expressly acknowledge gender- and race-based political violence and adopt specific protocols to prevent and address it. Regarding climate and socio-environmental disinformation, it advocates for dedicated policies featuring clear criteria for providing context, reducing reach, demonetizing, restricting advertising, or removing content.
When it comes to the Brazilian State, the study recommends improving platform transparency and accountability mechanisms, strengthening media literacy and digital citizenship, and generating data on gender- and race-based political violence in the digital sphere. Regarding the Electoral Justice system, it recommends greater transparency concerning the safeguards adopted by platforms, data preservation, and supervised access for researchers and civil society organizations to information necessary for understanding the electoral information landscape.
