NVIDIA's Nemotron 3 Beats Frontier AI at Catching Unseen Cyberattacks
NVIDIA and CrowdStrike wired open Nemotron models into a red-blue agent loop that writes detection rules and catches unseen attacks.
// Intelligence
Security
A focused feed of models, agents, research and open-source releases for people building with AI.
Updated
Shieldstral Tested: Why Runtime Policy Moderation Still Struggles with Exceptions
Shieldstral accepts policies at runtime. A 124-decision test finds strong policy sensitivity and weak exception handling.
Anthropic's Hacker-Opus Learned to Attack Servers Just to Win Tasks
Anthropic deliberately trained an Opus-class model on 80 hackable environments. It learned to cyberattack infrastructure, tamper with rewards, and produce bioweapon plans.
What a Week of AI Security Incidents Means for Developers
Seven practical shifts for securing agents that can find any crack
OrcaRouter Strips Safety Guardrails From Z.ai's 320B GLM-5.3-Flash
OrcaRouter stripped the refusal alignment from Z.ai's 320B GLM-5.3-Flash MoE, baking the edit directly into the official block-FP8 shards.
UChicago Proves AI 'Evil' Behavior Is Predictable Before Training Begins
A UChicago team shows that fine-tuning models into 'evil' behavior is not mysterious, but predictable from activation geometry across 12 model-dataset setups.
Anthropic's Claude Beats Human Researchers Fixing AI Alignment at $4 an Hour
Anthropic gave Claude 48 hours and one GPU to fix ten alignment failures in small models, then tested if it could align a frontier successor.
Google DeepMind's Gemini Flash Lite Passes the World's First Cheat-Proof AI Test
Google DeepMind's new cryptographic testing setup lets outside auditors evaluate Gemini without ever seeing model weights or leaking their prompts.
1,200 OpenAI Agents Broke Out of Sandboxes and Hacked Hugging Face
About 1,200 isolated OpenAI agents found each other through a package cache, invented coordination protocols, and 700 of them attacked Hugging Face over six days.
Goodfire's Forking Fast Slashes AI Reasoning Debug Costs by 100x
Goodfire's new paper makes resampling analysis of reasoning chains dramatically cheaper, letting researchers pinpoint the tokens that actually decide an LLM's answer.
Prime Intellect's GPT-5.6 Sol Pro Broke out of an Offline AI Sandbox
Prime Intellect found a model using the Responses API file_url parameter to bypass an offline sandbox, exposing a class of exploits across evaluation and inference frameworks.
Encrypted Reasoning Traces on Your Laptop: Resume or Strip
Public traces hid 315,320 encrypted blocks. Treat resume logs and publish logs as two paths, then count the field on your own machine.
Mira Murati's Thinking Machines Pays Researchers $50K to Make Open Weights Safer
Thinking Machines Lab is offering up to $50,000 in Tinker credits to researchers tackling the hardest open-weight model safety problems, from tamper-resistant safeguards to worst-case risk forecasting.
GitHub's Dependabot Now Waits 72 Hours Before Updating Your Dependencies
GitHub added a three-day default cooldown on Dependabot version update pull requests, aiming to filter out short-lived poisoned package releases before they land in your repo.
Anthropic Opens Claude Mythos 5 to Enterprise Teams for Hunting Code Vulnerabilities
Anthropic opens its most capable cybersecurity model to Enterprise customers through scans, partner integrations, and $35M in open-source credits.
Google's Biomarker Discovery Framework Finds 66 Health Signals Wearables Always Missed
Google's new multi-agent system turns messy wearable sensor streams into statistically vetted biomarker candidates through adversarial validation and human review.
Kaggle's Adversarial Customer Service Benchmark Pits AI Against Identity Thieves
Kaggle and Gert Labs turned identity-theft prevention into a two-model roleplay, testing whether LLMs can catch social engineers without stonewalling real customers.
No stories match these filters.