Brazil Preemptively Bans Deepfakes in Elections, Raising Concerns Over Democratic Process
Brazil’s Superior Electoral Court (TSE) has taken a decisive step to combat the potential spread of misinformation during this year’s municipal elections by prohibiting the use of deepfakes and imposing restrictions on artificial intelligence (AI) in political campaigns. While the move has been lauded by some as a proactive measure against the rising threat of AI-generated fake content, the method by which the court enacted the resolution has sparked debate about its adherence to democratic norms. Minister Alexandre de Moraes introduced the legislation to Congress but ultimately bypassed a Congressional vote, opting instead to implement the resolution solely through the judiciary. This decision, seemingly driven by concerns about the potential disruptive power of AI, raises questions about the balance between safeguarding elections and upholding established legislative processes.
The Rise of AI-Generated Misinformation: Examining the Evidence
An analysis of AI-related fact-checks aggregated by Google over the past year reveals a notable increase in Portuguese-language fact-checks addressing AI-generated content. However, the content flagged during this period appears to be relatively benign, at least for now. The majority of the viral AI-generated material fact-checked between March 2023 and February 2024 focused on topics outside of Brazil and held little relevance to past or present Brazilian elections. This raises the question of whether the TSE’s sweeping ban, while well-intentioned, is proportionate to the current threat level. The situation in Brazil highlights the challenging balancing act faced by lawmakers worldwide – the need to act swiftly to preempt potential dangers while ensuring that such actions are grounded in evidence and implemented through appropriate democratic channels.
Delving Deeper: A Quantitative Analysis of AI-Related Fact-Checks
A closer examination of Google’s Fact Check Tool API, which compiles fact-checks from professional organizations, provides further insight into the scope of the issue. Between March 2023 and February 2024, Google indexed 509 fact-checks mentioning "artificial intelligence" or "deepfake" in the ten most spoken languages globally. This translates to an average of 42 fact-checks per month addressing AI-related falsehoods, distortions, or simply discussing the topic. Significantly, this number is on the rise. In the first two months of 2024 alone, the average number of AI-related fact-checks jumped to 70 per month, a 62% increase compared to 2023. This surge suggests a growing awareness of and concern about the potential for AI to be misused for disinformation purposes.
Portuguese-Language Fact-Checks: A Focus on Brazil
Drilling down into the Portuguese-language data, 25 fact-checks addressing AI-generated falsehoods or mentioning AI were published during the 12-month period. This number also reflects the upward trend observed across other languages. The monthly average of AI-related fact-checks in Portuguese rose from 1.8 in 2023 to 3.5 in the first two months of 2024. Nine media organizations, seven Brazilian and two Portuguese, contributed to these fact-checks, with the majority adhering to the International Fact-Checking Network’s Code of Ethics. This indicates a growing commitment within the Portuguese-speaking fact-checking community to address the issue of AI-generated misinformation.
The Brazilian Context: Limited Impact of AI-Generated Disinformation So Far
Focusing specifically on Brazil, 19 articles were published in Portuguese during the analyzed period flagging controversial AI use or mentioning the topic. Notably, almost a third of these fact-checks emerged in 2024, reinforcing the upward trend. However, crucially, none of the fact-checked content related to national politics or politicians. No political parties or coalitions were mentioned in the AI-fabricated pieces. The only content mentioning elections addressed a manipulated video of World Economic Forum founder Klaus Schwab, falsely portraying him as advocating for AI to replace electoral processes. This suggests that, despite the TSE’s concerns, AI-generated disinformation has not yet significantly impacted Brazilian politics.
Analyzing the Content: Trivial Falsehoods and Lack of Political Impact
The content analysis of the fact-checked material reveals a striking lack of impactful or significant AI-generated falsehoods capable of influencing public policy or electoral outcomes. The identified examples range from a fabricated "demonic park" to a fake photo of Julian Assange in prison and the alleged launch of an "artificial woman" in China. The fact-checks involving the term "government" did not involve manipulated content but rather addressed misinformation about the World Economic Forum and a manipulated video featuring a television presenter unrelated to politics. This analysis reinforces the observation that while AI-generated disinformation is present, its impact in Brazil has thus far been limited to relatively trivial matters. This raises questions about whether the TSE’s preemptive ban, while potentially justifiable in the long term, is a proportionate response to the current, relatively low-level threat posed by AI-generated misinformation in the Brazilian political landscape.
