AI Governance

Zero-Knowledge Proofs for Frontier AI Training Verification

New research demonstrates that zero-knowledge proofs can verify frontier AI training claims without exposing model weights or data—a breakthrough for AI governance, authenticity, and trust in synthetic media systems.

model collapse

Model Collapse as Epidemic: SIR Dynamics for Synthetic Data

A new arXiv paper frames model collapse from synthetic data contamination as an epidemiological problem, applying bilayer SIR dynamics to model how AI-generated content spreads through training corpora and degrades future models.

deepfakes

Underground Markets Now Sell Deepfake Kits and AI Bots

Criminal marketplaces are increasingly trading turnkey deepfake kits, voice-cloning tools, and AI-powered hacking bots, lowering the barrier for fraud, impersonation scams, and synthetic identity attacks at unprecedented scale.

deepfake detection

Reality Defender Flags Deepfake Policy and Detection Gaps

Reality Defender outlines critical gaps in deepfake policy, details its multi-model detection approach, and unveils a new integration aimed at expanding real-time synthetic media defense across enterprise communication channels.

Netflix

Netflix Bets on Voice AI and Generative Tech for Discovery

Netflix is exploring voice AI and generative technology to combat 'content overload,' signaling deeper integration of synthetic media tools into its discovery and recommendation experience.

LLM Infrastructure

Breaking the KV Wall: Scaling LLM Inference at Scale

As LLMs handle longer contexts and more concurrent users, the KV cache has become the dominant bottleneck in inference. New architectural approaches aim to break through this memory wall for next-generation serving.

LLM

Why the Last Token Isn't Your LLM's Best Embedding

New analysis challenges the standard practice of extracting embeddings from the final token of decoder-only LLMs, showing intermediate layers and alternative pooling strategies often produce richer semantic representations.

TSMC

TSMC CEO: AI Chip Demand Won't Be Met for Years

TSMC CEO C.C. Wei says it will be a long time before the foundry can fully satisfy AI chip demand, signaling prolonged supply constraints for Nvidia, AMD, and the broader compute stack powering generative AI.

Quantization

Recover-LoRA: Restoring Accuracy in 2-Bit LLMs

A new technique called Recover-LoRA uses low-rank adaptation and knowledge distillation on synthetic data to reclaim accuracy lost during aggressive 2-bit quantization of language models, enabling far more efficient deployment.

AI Licensing

Hasbro Opens AI Licensing Program for Iconic Characters

Hasbro is launching a formal licensing program to bring characters like Transformers, My Little Pony, and Dungeons & Dragons into AI-powered experiences, signaling a major shift in how IP holders engage with generative AI platforms.

LLM Evaluation

LLM-as-Judge Geometry: Consensus Isn't Human Alignment

New research challenges the assumption that agreement between LLM judges signals human alignment. The geometry of LLM evaluation reveals systematic biases that affect how synthetic content and AI outputs are assessed.

deepfake detection

Android Now Detects Deepfake Scam Calls Faking Contacts

Google rolls out new Android protections that flag voice-cloned scam calls impersonating your contacts, bringing on-device deepfake detection to billions of phones and raising the bar for synthetic audio defenses.