Yazılar

Meta to Require AI Disclosure for Political Ads Ahead of Canadian Elections

/in /tarafından

Meta Platforms (META.O) announced on Thursday that it will require advertisers to disclose the use of AI or other digital techniques in political or social issue ads ahead of Canada’s federal elections. This move aims to combat misinformation and increase transparency in the political advertising landscape.

The new disclosure rule will apply to ads featuring photorealistic images, videos, or realistic-sounding audio that have been digitally altered to show a real person saying or doing something they did not actually say or do. It will also apply to ads showcasing non-existent individuals or fabricated events, altered footage of real events, or misleading depictions of events that may not be accurate.

In November 2023, Meta extended its ban on new political ads following the U.S. election to combat misinformation. The company also prohibited political campaigns and advertisers in regulated sectors from using its generative AI advertising tools. Despite these efforts, Meta had a setback earlier this year when it scrapped its U.S. fact-checking programs amid pressure from conservatives to overhaul its approach to political content.

Additionally, Meta has introduced a feature allowing users to disclose when they share AI-generated content, enabling the platform to label such media accordingly.

Meta to Launch “Community Notes” in the U.S. Using X’s Algorithm

/in /tarafından

Meta will begin testing its new Community Notes feature in the U.S. starting March 18, utilizing technology from Elon Musk’s X, the company announced on Thursday. This move comes two months after Meta scrapped its fact-checking program under pressure from conservatives, signaling a shift from traditional fact-checking to a crowd-sourced model.

The feature will allow users to write and rate notes to flag false or misleading content across Instagram, Facebook, and Threads, effectively replacing the third-party fact-checkers that were previously responsible for content moderation. 200,000 U.S. users have already signed up as potential contributors to the new system.

Meta’s switch to the Community Notes program represents a significant overhaul in its approach to content management. The company has been keen to improve its relationship with the Trump administration, which has criticized social media platforms for silencing conservative voices. President Donald Trump praised Meta’s decision in January, acknowledging the shift toward a more inclusive and less biased content moderation process.

To power Community Notes, Meta will adopt X’s open-source algorithm, which was originally developed as part of X’s Birdwatch feature. The system, now known as Community Notes, allows users to contribute and vote on content’s accuracy. Meta’s version will limit notes to 500 characters and initially support six languages: English, Spanish, Chinese, Vietnamese, French, and Portuguese. Notes will remain anonymous and will be published only if users with differing viewpoints agree that the note provides helpful context.

Contributors must be over 18 and include a supporting link when posting notes. Meta has emphasized that this system will be less biased than the previous third-party fact-checking method. Once the new system is in place, third-party fact-check labels will no longer appear on U.S. content.

Meta, which boasts over 3 billion global users, continues to collaborate with nearly 100 certified fact-checking organizations across 60+ languages, according to the company.

Spain Moves to Fine Companies for Unlabelled AI-Generated Content

/in /tarafından

Spain’s government has approved a new bill imposing hefty fines on companies that fail to label AI-generated content properly. The measure, aimed at combating misinformation and the spread of deepfakes, aligns with the European Union’s AI Act, which enforces strict transparency rules for high-risk AI applications.

Digital Transformation Minister Oscar Lopez emphasized the dual nature of AI, describing it as both a powerful tool for improving lives and a potential threat to democracy through disinformation. Spain is among the first EU nations to implement these regulations, setting a more rigid standard compared to the United States’ largely voluntary approach.

The proposed law classifies the failure to properly label AI-generated content as a “serious offense,” punishable by fines of up to €35 million ($38.2 million) or 7% of a company’s global annual revenue. The bill also prohibits subliminal AI techniques used to manipulate vulnerable populations, such as chatbots that encourage gambling addiction or AI-powered toys that promote risky behavior among children.

Another key provision bans the use of AI to classify individuals based on biometric data for scoring purposes, preventing organizations from assessing a person’s eligibility for benefits or predicting criminal behavior. However, authorities will still be permitted to use real-time biometric surveillance for national security purposes.

Spain’s newly established AI supervisory agency, AESIA, will oversee enforcement, except in areas such as data privacy, elections, finance, and crime, which will fall under their respective regulatory bodies. The bill must still pass the lower house before becoming law.