deepfakes
New research examines sexualized deepfake abuse from both perpetrator and victim perspectives, revealing psychological impacts and motivations behind this growing form of synthetic media exploitation.
LLM Security
Data poisoning attacks targeting large language models can manipulate outputs by corrupting training datasets. Understanding these vulnerabilities is critical for maintaining AI system integrity and authenticity.
AI music
New Deezer survey shows 97% of listeners struggle to distinguish AI-generated music from human compositions, highlighting detection challenges in synthetic audio as platforms grapple with authentication.
deepfakes
New survey reveals 33% of enterprises lack confidence in detecting deepfake attacks, highlighting critical gap in organizational security as synthetic media threats escalate across corporate environments.
deepfakes
Sumsub reports sophisticated fraud including deepfake attacks increased 180% globally in 2024, with UK deepfake incidents doubling. Identity verification systems face escalating synthetic media threats as fraudsters deploy AI-generated content.
AI video generation
Professional video creators share insights on integrating AI video generation into workflows, revealing both the limitations of current synthetic media tools and emerging use cases in the creative industry.
AI Deception
New research presents HUMA, an AI facilitator engineered to pass as human in multi-user conversations. The system raises critical questions about digital authenticity and AI deception in social contexts.
AI detection
New research reveals AI detection tools incorrectly classify lightly polished Arabic articles as AI-generated, exposing critical vulnerabilities in content authentication systems that impact non-English text disproportionately.
TikTok
TikTok launches new preference controls allowing users to adjust how much AI-generated content appears in their feeds, building on its AI labeling system to give users more transparency and control over synthetic media exposure.
Computer Vision
AI doesn't perceive images like humans do. Understanding tokenization, embeddings, and attention mechanisms reveals how neural networks process visual data—critical knowledge for developing and detecting synthetic media.
AI Transparency
Researchers propose comprehensive workflow for tracking AI decision-making processes from data input to final output, addressing transparency and accountability challenges in modern AI systems through systematic documentation and verification methods.
LLM Watermarking
Researchers introduce WaterMod, a modular token-rank partitioning approach that improves LLM watermarking by maintaining probability balance across model outputs, enhancing detection while preserving text quality.