OpenAI's Sora video tool challenges traditional media creation
- Meta has developed a new tool called Meta Video Seal to combat deepfakes.
- The tool is open source and can withstand common video edits and compression.
- This initiative aims to raise awareness about watermarking in AI-generated content.
In December 2024, Meta announced the release of a new tool called Meta Video Seal, aimed at combating the rising issue of deepfakes that have been reported to comprise 7% of all fraud cases that year. This tool, which is open source and can be integrated into existing software, applies imperceptible watermarks to AI-generated video clips, thereby protecting originality and aiding in the detection of manipulated content. Meta's move comes as deepfakes have increasingly been used in impersonations and social engineering scams, raising concerns about privacy and security. The Meta Video Seal is designed to withstand common video editing techniques and compression methods that might typically compromise watermark visibility. Pierre Fernandez, an AI research scientist at Meta, highlighted the lack of robustness in existing watermarking solutions and the need for more effective technology. Additional features include the ability to embed a hidden message in videos that can later be uncovered for origin verification. To encourage the adoption of such watermarking techniques, Meta is launching the Meta Omni Seal Bench, a public leaderboard to compare various watermarking methods, and organizing workshops to foster collaboration within the academic and industry communities. This initiative signifies a potential collaborative future for addressing the challenges posed by AI-generated content, although Fernandez acknowledges that some developers may prefer existing proprietary solutions unless clear incentives arise for integration.