Fastino Labs has released GLiGuard, a 300M parameter open-source safety moderation model that evaluates four safety tasks — prompt safety, jailbreak strategy detection, harm category classification, and refusal detection — in a single forward pass. Built on an encoder architecture rather than the decoder-only design used by most guardrail models, GLiGuard achieves up to 16x higher throughput and 16.6x lower latency than current state-of-the-art models, while matching or exceeding the accuracy of models 23 to 90 times its size across nine safety benchmarks. Model weights are available under the Apache 2.0 license on Hugging Face.

The post Fastino Labs Open-Sources GLiGuard: A 300M Parameter Safety Moderation Model That Matches or Exceeds Accuracy of Models 23–90x Its Size appeared first on MarkTechPost.

63% of organizations have no AI governance policy. Shadow AI is already running inside your stack — here is the data.

The post Enterprise AI Governance in 2026: Why the Tools Employees Use Are Ahead of the Policies That Cover Them appeared first on MarkTechPost.

In this tutorial, we build a realistic Zero-Trust network simulation by modeling a micro-segmented environment as a directed graph and forcing every request to earn access through continuous verification. We implement a dynamic policy engine that blends ABAC-style permissions with device posture, MFA, path reachability, zone sensitivity, and live risk signals such as anomaly and […]

The post How to Build a Dynamic Zero-Trust Network Simulation with Graph-Based Micro-Segmentation, Adaptive Policy Engine, and Insider Threat Detection appeared first on MarkTechPost.

Nous Research releases Token Superposition Training (TST), a two-phase pre-training method that cuts wall-clock training time by up to 2.5x at matched FLOPs by averaging contiguous token embeddings into bags during Phase 1 and reverting to standard next-token prediction in Phase 2 — without changing the model architecture, tokenizer, optimizer, or inference-time behavior. Validated at 270M, 600M, 3B dense, and 10B-A1B MoE scales.

The post Nous Research Releases Token Superposition Training to Speed Up LLM Pre-Training by Up to 2.5x Across 270M to 10B Parameter Models appeared first on MarkTechPost.

Google DeepMind researchers have outlined four interaction principles and released experimental demos of an AI-enabled mouse pointer powered by Gemini — one that captures the visual and semantic context around the cursor so users can point, speak in natural shorthand, and get things done without switching to a separate AI window.

The post Google DeepMind Introduces an AI-Enabled Mouse Pointer Powered by Gemini That Captures Visual and Semantic Context Around the Cursor appeared first on MarkTechPost.

Thinking Machines Lab has introduced a research preview of TML-Interaction-Small, a 276B parameter Mixture-of-Experts model with 12B active parameters, built around a multi-stream, time-aligned micro-turn architecture that processes 200ms chunks of audio, video, and text simultaneously — eliminating the need for external voice-activity detection harnesses. Unlike standard turn-based models that freeze perception during generation, the system runs two components in parallel: a real-time interaction model that maintains continuous full-duplex exchange with the user, and an asynchronous background model that handles sustained reasoning and tool use while sharing the full conversation context throughout.

The post Mira Murati’s Thinking Machines Lab Introduces Interaction Models: A Native Multimodal Architecture for Real-Time Human-AI Collaboration appeared first on MarkTechPost.